Automated conversational survey system and survey method using same

Abstract

An automated conversational survey system according to one embodiment of the present invention comprises: a survey input unit that receives content in accordance with a form of a survey; a preprocessing unit connected to the survey input unit to analyze the input content for each survey item and sequentially generate questions; an output unit connected to the preprocessing unit and outputting voice or text for each survey item; a response input unit that receives voice or text of a survey respondent; a determination control unit connected to the preprocessing unit and the response input unit, and determining a situation for received response content and classifying and responding to same according to a scenario; and a storage unit connected to the determination control unit to store survey content information.

Description

Automated interactive survey system and survey method using the same

[0001] The present invention relates to an automatic interactive survey system and a survey method using the same, and more particularly, to an automatic interactive survey system and a survey method using the same that can efficiently conduct surveys targeting people who have difficulty responding to surveys using artificial intelligence technology.

[0002] A survey is a research method that collects and analyzes data on social phenomena through pre-structured questionnaires or interviews. The purpose of such surveys is to gather information from a large group of respondents presumed to represent a certain population. While paper questionnaires were traditionally used, recent advances in information and communication technologies like the Internet and smartphones have led to the rise of online survey methods.

[0003] The survey results derived from these surveys must be guaranteed to be reliable, but their reliability is reduced by various factors, and is particularly greatly affected by the quality of the data required to derive the results.

[0004] Traditional surveys involve researchers reading questions directly to respondents and recording their responses. This process is time-consuming and labor-intensive. This can be particularly challenging for the elderly, children, and those with limited mobility.

[0005] For example, if a subject responds insincerely to a survey and gives answers unrelated to the survey content, or if response data provided for malicious purposes is reflected in the survey results, there is a problem that the results derived from only normal response data may be distorted.

[0006] The present invention has been devised to solve the above problems, and the purpose of the present invention is to provide an automatic interactive survey system that can efficiently conduct surveys targeting people who have difficulty responding to surveys using artificial intelligence technology.

[0007] Another object of the present invention is to provide a survey method using an automatic interactive survey system that can efficiently conduct surveys targeting people who have difficulty responding to surveys using artificial intelligence technology.

[0008] An automatic interactive survey system according to one embodiment of the present invention comprises: a survey input unit for receiving content in a survey format; a preprocessing unit connected to the survey input unit for analyzing the input content by item and generating questions in order; an output unit connected to the preprocessing unit for outputting voice or text by item; a response input unit for receiving voice or text from a survey subject; a judgment control unit connected to the preprocessing unit and the response input unit for judging a situation with respect to the received response content and classifying and responding according to a scenario; and a storage unit connected to the judgment control unit for storing survey content information.

[0009] In an automatic interactive survey system according to one embodiment of the present invention, the preprocessing unit is characterized in that it is linked to an external large language model (LLM) module via an application programming interface (API).

[0010] In an automatic interactive survey system according to one embodiment of the present invention, the judgment control unit is characterized in that it judges the situation based on the received response content by linking to an external LLM module via an API.

[0011]

[0012] In addition, a survey method according to another embodiment of the present invention includes: (A) a step in which a judgment control unit uploads the survey content together with the survey form when the survey content is entered in a survey form through a survey input unit; (B) a step in which a preprocessing unit links to an external large language model (LLM) module via an application programming interface (API) to analyze the survey content and sequentially generate questions; (C) a step in which an output unit outputs the received question to a respondent in voice or text form under the control of the judgment control unit; (D) a step in which a response input unit collects a survey subject's voice response and a text of a response message for the output question; and (E) a step in which the judgment control unit links to an external LLM module via an API to analyze the collected responses and determine their validity.

[0013] A survey method according to another embodiment of the present invention is characterized in that (F) the judgment control unit further includes a step of outputting additional questions to the respondent through the preprocessing unit and the output unit when the collected responses are judged to be inappropriate responses.

[0014] A survey method according to another embodiment of the present invention is characterized in that (G) the judgment control unit further includes a step of storing the collected responses in a storage unit when the judgment control unit determines that the collected responses are valid response contents for each scenario.

[0015] In a survey method according to another embodiment of the present invention, step (B) is characterized in that the preprocessing unit generates an implicit question by combining similar questions among the questions of the input survey content in conjunction with the LLM module.

[0016]

[0017] The features and advantages of the present invention will become more apparent from the following detailed description based on the attached drawings.

[0018] Prior to this, the terms and words used in this specification and claims should not be interpreted in their usual or dictionary meanings, but should be interpreted in the sense and concept that is consistent with the technical idea of ​​the present invention based on the principle that the inventor can appropriately define the concept of the term to explain his or her own invention in the best way.

[0019] A survey method using an automatic interactive survey system according to an embodiment of the present invention uses artificial intelligence to conduct a convenient and efficient survey targeting people who have difficulty responding to surveys, and has the effect of saving time and cost.

[0020] The automatic interactive survey method according to an embodiment of the present invention has the effect of improving the reliability of data by increasing the accuracy and consistency of responses.

[0021] Figure 1 is a configuration diagram of an automatic interactive survey system according to an embodiment of the present invention.

[0022] Figure 2 is a flowchart for explaining a survey method using an automatic interactive survey system according to an embodiment of the present invention.

[0023]

[0024] The purpose, specific advantages, and novel features of the present invention will become more apparent from the following detailed description and preferred embodiments in conjunction with the accompanying drawings. In this specification, when assigning reference numbers to components in each drawing, it should be noted that, as much as possible, identical components are given the same numbers even if they are shown in different drawings. In addition, while terms such as first, second, etc. may be used to describe various components, the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. In addition, in describing the present invention, if a detailed description of a related known technology is judged to unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

[0025] Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to the attached drawings. Figure 1 is a configuration diagram of an automatic interactive survey system according to an embodiment of the present invention.

[0026] The automatic interactive survey system according to an embodiment of the present invention is an automatic interactive survey system using artificial intelligence that can efficiently conduct surveys on people who have difficulty responding to surveys, such as the elderly, children, and people with disabilities with limited movement.

[0027] Specifically, an automatic interactive survey system according to an embodiment of the present invention includes a survey input unit (110) for receiving content in a survey format, a preprocessing unit (120) connected to the survey input unit (110) for analyzing the input content by item and sequentially creating questions, an output unit (130) connected to the preprocessing unit (120) for outputting voice or text by item, a response input unit (140) for receiving voice or text from a survey subject, a judgment control unit (150) for judging a situation with respect to the received response content and classifying and responding according to a scenario, and a storage unit (160) connected to the judgment control unit (150) for storing survey content information.

[0028] The survey input unit (110) receives and processes the structure and question list of a survey written by a user in a document format such as Hangul, MS Word, or Excel. Users can input the survey content as text or upload an existing survey form. The survey content entered in this manner is processed by the survey input unit (110) and transmitted to the preprocessing unit (120).

[0029] The preprocessing unit (120) is a part that connects to an external LLM (large language model) module via an API (application programming interface), analyzes the input survey content by question and sequentially generates questions. In this process, it connects to a fine-tuned LLM module to analyze the input text, and the LLM module is trained appropriately for each conversation channel and independently separates each question. The questions preprocessed in this way are transmitted to the output unit (130).

[0030] Here, the LLM module is a module that includes LLM, which consists of an artificial neural network with a large number of parameters (usually billions of weights or more), and trains to learn a significant amount of unlabeled text using self-supervised learning or semi-self-supervised learning.

[0031] These LLM modules are built using deep learning techniques to process massive amounts of natural language data and generate questionnaires that are indistinguishable from human-generated text. They can learn from vast amounts of text data and derive responses based on user requests. These LLMs include OpenAI's Generative Pre-trained Transformer (GPT) series and Google's Bidirectional Encoder Representations from Transformers (BERT) models. These models are being used in a variety of applications, including language translation, content generation, and chatbots.

[0032] In particular, the preprocessing unit (120) can analyze the input survey content by question in conjunction with the LLM module and combine similar questions to generate implicit questions. For example, among the survey questions, questions such as "area of ​​residence," "neighborhood where you live," and "township / village where you currently live" can be summarized and generated as the question "What is your current address?"

[0033] The output unit (130) includes a display device or speaker connected to the preprocessing unit (120), and can output the contents of the items transmitted from the preprocessing unit (120) in voice or text under the control of the judgment control unit (150).

[0034] The response input unit (140) includes a microphone that receives the voice of the survey subject responding to the question output through the output unit (130) or a display device that receives text, and transmits the received voice or text information to the judgment control unit (150).

[0035] The judgment control unit (150) controls the overall automatic conversational survey system, judges the situation based on the responses received from the response input unit (140), and classifies and responds according to the scenario. This judgment control unit (150) is a unit that connects to an external LLM module via an API, and can create an artificial intelligence chatbot capable of voice recognition based on a database by connecting to this LLM module. The artificial intelligence chatbot can utilize NLP (Natural Language Processing), NLU (Natural Language Understanding), and NLG (Natural Language Generation). At this time, NLU extracts keywords and related words and extracts patterns through connection with a database, and NLG can create and respond to rule-based conversations by analyzing sentence structures through connection with a database.

[0036] Specifically, the AI ​​chatbot generated in this way can preprocess, interpret, and output results corresponding to the intention of the natural language included in at least one scenario. The main task performed in the natural language processing stage of the AI ​​chatbot is morphological analysis of natural language. Morphological analysis (POS-Tagging) refers to the task of dividing the raw corpus into morpheme units and attaching part-of-speech information to each form. It can identify the structure of various linguistic properties such as morphemes, roots, prefixes / suffixes, and parts-of-speech (POS), so that it can respond to the response content using the chatbot according to the scenario.

[0037] And, in the Natural Language Understanding stage of the AI ​​chatbot, the respondent's response can be decomposed and then the appropriate syntax and semantic system can be determined for the decomposed entities. At this time, the Natural Language Understanding stage of the AI ​​chatbot can perform processing such as general named entity recognition, detailed named entity recognition, lexical semantic analysis, syntax analysis, and intent analysis. When extracting named entities, a detailed named entity recognizer (Specific-NER) can be further utilized for the chatbot to operate in a specific domain or task, and for named entity recognition or semantic role determination, not only statistical modeling such as Hidden Markov Model, Support Vector Machines, and Conditional Random Fields, but also deep learning such as Feed-Forward Neural Network (FFNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) can be used.

[0038] The storage unit (160) organizes and stores the responses received according to the questionnaire content under the control of the judgment control unit (150). All response data is safely stored in the storage unit (160) and can be used for subsequent analysis and report creation. The storage unit (160) can also perform regular backups to maintain data integrity.

[0039] The automatic interactive survey system according to the embodiment of the present invention configured in this way can efficiently conduct surveys using artificial intelligence targeting people who have difficulty responding to surveys, such as the elderly, children, and people with disabilities with limited movement.

[0040]

[0041] Hereinafter, a survey method using an automatic interactive survey system according to an embodiment of the present invention will be described with reference to FIG. 2. FIG. 2 is a flowchart illustrating a survey method using an automatic interactive survey system according to an embodiment of the present invention.

[0042] In a survey method using an automatic interactive survey system according to an embodiment of the present invention, first, a user inputs the survey content in a survey form through a survey input unit (110), and then the judgment control unit (150) uploads the survey content together with the survey form (S210).

[0043] The survey input unit (110) receives and processes the structure and list of questions of a survey written by a user in a document format such as Hangul, MS Word, or Excel. The user can input the contents of the survey as text or upload an existing survey form. The contents of the survey thus entered are processed by the survey input unit (110) under the control of the judgment control unit (150) and transmitted to the preprocessing unit (120).

[0044] After uploading the survey content, the preprocessing unit (120) analyzes the survey content and sequentially generates questions by linking to an external LLM (large language model) module via an API (application programming interface) (S220).

[0045] That is, the preprocessing unit (120) analyzes the input text in conjunction with a fine-tuned LLM module, and the LLM module is trained appropriately for each conversation channel to independently separate and analyze each question, and can transmit the preprocessed questions to the output unit (130).

[0046] In particular, the preprocessing unit (120) can analyze the input survey content by question in conjunction with the LLM module and combine similar questions to generate implicit questions. For example, among the survey questions, questions such as "area of ​​residence," "neighborhood where you live," and "township / village where you currently live" can be summarized and generated as the question "What is your current address?"

[0047] Thereafter, the output unit (130) can output the received preprocessed questions to the respondent in voice or text form under the control of the judgment control unit (150) (S230).

[0048] The output unit (130) includes a display device or a speaker, and can output the question content transmitted from the preprocessing unit (120) to the respondent in voice or text form under the control of the judgment control unit (150).

[0049] For the question content printed in this way, the response input unit (140) collects the response voice and text of the response message of the survey subject (S240).

[0050] To this end, the response input unit (140) includes a microphone that receives the voice of the survey subject responding or a display device that receives the text of the response message, and transmits the received voice or text information to the judgment control unit (150).

[0051] The judgment control unit (150) analyzes the collected responses by linking to an external LLM module via API and determines their validity (S250).

[0052] Specifically, the judgment control unit (150) can generate an artificial intelligence chatbot capable of voice recognition based on a database by linking with the LLM module, and can preprocess, interpret, and output results corresponding to the intention of the natural language contained in the collected responses. At this time, the main work performed in the Natural Language Process stage is to analyze the morphemes of the natural language, and in the Natural Language Understanding stage, after decomposing the respondent's response, determine an appropriate syntax and semantic system for the decomposed entities, and perform processing such as general entity recognition, detailed entity recognition, lexical semantic analysis, syntax analysis, and intention analysis to determine the validity for each scenario.

[0053] Accordingly, the judgment control unit (150) executes processing to output additional questions to the respondent through the preprocessing unit (120) and the output unit (130) if the collected response is inappropriate (S260).

[0054] On the other hand, the judgment control unit (150) stores the response information in the storage unit (160) if the collected response is a valid response content for each scenario (S270).

[0055] For example, in the first scenario, if the respondent gives an appropriate response to the questionnaire, the judgment control unit (150) determines that it is a valid response and stores the response information in the storage unit (160).

[0056] In the second scenario, if the respondent does not respond to the questionnaire within a specified time and does not respond, the judgment control unit (150) can output a prompting message or voice such as “Could you please say that again?” to the respondent through the output unit (130).

[0057] If there is still no response, the system repeats the prompt message or voice output according to the set waiting number or time, and then moves on to the next question. At this time, the response information is recorded and stored as “no response” in the storage unit (160).

[0058] In the third scenario, if the respondent gives an incomplete or inappropriate response, the judgment control unit (150) can output an additional question to the respondent through the output unit (130) in the form of a prompt message or voice, such as, "Could you please answer again clearly?" or "That answer is incomplete. Please tell me more about ~." In this case, the judgment control unit (150) specifically explains which part is incomplete and requests an explanation thereof, and if the response is still repeatedly incomplete, the response information is recorded and stored in the storage unit (160) as is.

[0059] In the fourth scenario, if the respondent's response is judged to be incomprehensible, the judgment control unit (150) classifies it as an incomprehensible response and outputs a message or voice such as "I'm sorry. That's an incomprehensible answer. Could you please say that again?" to the respondent through the output unit (130) to repeatedly perform the re-questioning. At this time, the judgment control unit (150) repeats the re-questioning according to the set number of times, and if it is judged to be a repetitive incomprehensible response, the response information is recorded and stored in the storage unit (160).

[0060] In the fifth scenario, if the respondent misunderstands the question and the response is judged to be a misunderstanding, the decision control unit (150) can output to the respondent through the output unit (130) a message or voice message such as "I will ask the question again in a different way" while explaining the question in a different way or giving an example. At this time, the decision control unit (150) can appropriately reconstruct the explanation of the question based on the respondent's age, gender, and other information and circumstances given in advance, and output it to the respondent through the output unit (130). However, if the response is judged to be a repetitive misunderstanding, the decision control unit (150) records and stores the response information as is in the storage unit (160).

[0061] In the sixth scenario, when the respondent expresses a refusal to answer a question, the decision control unit (150) classifies it as a refusal to answer and outputs to the respondent via the output unit (130) a message or voice message stating, "I will skip this question and move on to the next question." Subsequently, the decision control unit (150) stores the response information in the storage unit (160) as a "refusal to answer" and moves on to the next question.

[0062] The survey method using an automated interactive survey system according to an embodiment of the present invention, which includes such a process, utilizes artificial intelligence to conduct convenient and efficient surveys targeting individuals who may otherwise struggle to respond, saving time and money. Furthermore, the automated interactive survey method according to an embodiment of the present invention can enhance data reliability by increasing the accuracy and consistency of responses.

[0063]

[0064] Although the technical idea of ​​the present invention has been specifically described in accordance with the above preferred embodiments, it should be noted that the above-described embodiments are for illustrative purposes only and are not intended to limit the invention.

[0065] In addition, a person skilled in the art will understand that various implementations are possible within the scope of the technical idea of ​​the present invention.

Claims

1. Survey input section for entering information in the form of a survey; A preprocessing unit that connects to the above survey input unit, analyzes the input contents for each item, and generates questions in order; An output unit connected to the above preprocessing unit and outputting voice or text for each item; A response input section that receives the voice or text of the survey subject; A judgment control unit connected to the above preprocessing unit and the response input unit, and judging the situation based on the received response content, classifying it according to the scenario, and responding; and A storage unit connected to the above judgment control unit and storing the survey content information; An automated interactive survey system including:

2. In paragraph 1, An automatic interactive survey system characterized in that the above preprocessing unit is linked to an external LLM (large language model) module via an API (application programming interface).

3. In paragraph 1, An automatic interactive survey system characterized in that the above judgment control unit judges the situation based on the received response content by linking to an external LLM module via an API. 4.(A) When the survey content is entered into the survey form through the survey input section, the judgment control section uploads the survey content along with the survey form; (B) A step in which the preprocessing unit analyzes the survey content and sequentially generates questions by linking to an external LLM (large language model) module via an API (application programming interface); (C) A step of outputting the question received through the output unit in voice or text form to the respondent under the control of the judgment control unit; (D) a step of collecting the voice response and text of the response message of the survey subject to the above question output by the response input section; and (E) A step in which the above judgment control unit analyzes the collected response and determines its validity by linking with an external LLM module via an API; A survey method including:

5. In paragraph 4, (F) A survey method characterized in that it further includes a step of outputting additional questions to the respondent through the preprocessing unit and the output unit when the judgment control unit determines that the collected response is inappropriate.

6. In paragraph 4, (G) A survey method characterized in that it further includes a step of storing the collected responses in a storage unit when the judgment control unit determines that the collected responses are valid response contents for each scenario.

7. In paragraph 4, The above step (B) is a survey method characterized in that the preprocessing unit generates an implicit question by combining similar questions among the questions in the input survey content in conjunction with the LLM module.

Images