Interactive teaching material device, interactive teaching material program, and interactive teaching material processing method

Abstract

The system can determine correctness even when learners' answers are not unique, enabling interactive learning. [Solution] The interactive learning material device conducts learning through dialogue with the learner via a terminal device. The interactive learning material device includes a control unit that performs the following processes: asking the learner questions related to understanding a predetermined problem (S501-S503), receiving answers from the learner to the questions (S504, S505), and using a generating AI to determine whether the learner's answer is correct, incorrect, or partially incorrect (S506, S507).

Description

Technical Field

[0001] This invention relates to an interactive teaching material device, an interactive teaching material program, and an interactive teaching material processing method for assisting learning.

Background Art

[0002] Conventionally, as learning via a terminal device such as a tablet, there is a technique of presenting a learning plan suitable for a learner through dialogue with the learner by posing a problem to the learner from the system side, accepting the learner's answer, and evaluating the answer (for example, refer to Patent Document 1 below).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the above-described conventional technology, in learning using a terminal device, when the answer to a problem is not unique, for example, in the case of a free description or an answer by speech other than options, it has been impossible to accurately determine the correctness. If it is an option, a unique answer (correct or incorrect) can be correctly determined, but when an answer by free description or speech is partially incorrect, it becomes impossible to correctly determine the correctness. In this case, interactive learning along the learning scenario cannot be continued.

[0005] An object of this invention is to be able to perform correct / incorrect determination and proceed with interactive learning even when the answer given by the learner is not unique, in order to solve the problems of the above-described conventional technology.

Means for Solving the Problems

[0006] To solve the above-mentioned problems and achieve the objective, the interactive teaching material device according to this invention is an interactive teaching material device that conducts learning through dialogue with a learner, and is characterized by comprising a control unit that performs the following processes: asking a predetermined question to the learner; receiving an answer to the question from the learner; and using a generating AI to determine whether the learner's answer is correct, incorrect, or partially incorrect.

[0007] Furthermore, the interactive teaching material device according to this invention is characterized in that the answer is one of the following: selection of a predetermined unique option, free-form text input, or dialogue using arbitrary speech, and the control unit makes a judgment on the answers in the free-form text and dialogue formats using the generating AI.

[0008] Furthermore, the interactive teaching material device according to this invention is characterized in that, in the judgment using the generating AI, the control unit individually assigns evaluation points to each correct element included in the answer, and makes a judgment in multiple stages, including correct, partially incorrect, and incorrect, based on the sum of the evaluation points of each correct element.

[0009] Furthermore, the interactive teaching material device according to this invention is characterized in that the control unit branches into scenarios based on the level of understanding of the question, based on the result of the determination.

[0010] Furthermore, the interactive teaching material device according to this invention is characterized in that the control unit includes, as information to be input to the generating AI, information on a method for determining the question, an output format after determination, and a plurality of pre-prepared answers to the question that are included in the scenario.

[0011] Furthermore, the interactive teaching material device according to this invention is characterized in that the control unit has a handwriting recognition engine that recognizes characters written in the free-writing format, and a speech recognition engine that recognizes speech spoken in the dialogue format.

[0012] Furthermore, the interactive teaching material device according to this invention is characterized in that the control unit includes a process for providing explanations to guide the learner to the correct answer according to the learner's level of understanding of the question.

[0013] Furthermore, the interactive teaching material program according to this invention is an interactive teaching material program that conducts learning through dialogue with a learner, and is characterized in that it causes a computer to perform the following processes: asking a predetermined question to the learner; receiving an answer to the question from the learner; and using a generating AI to determine whether the learner's answer is correct, incorrect, or partially incorrect.

[0014] Furthermore, the interactive teaching material method according to this invention is an interactive teaching material processing method in which learning is conducted through dialogue with a learner, characterized in that a computer performs the following processes: asking a predetermined question to the learner; receiving an answer from the learner to the question; and using a generating AI to determine whether the learner's answer is correct, incorrect, or partially incorrect. [Effects of the Invention]

[0015] The interactive teaching material device, interactive teaching material program, and interactive teaching material processing method according to this invention have the effect of being able to determine whether a learner's answer is correct or incorrect even when the answer is not unique, and to advance interactive learning. [Brief explanation of the drawing]

[0016] [Figure 1] Figure 1 is an explanatory diagram showing an example of a system configuration for an interactive teaching material device. [Figure 2A] Figure 2A is an explanatory diagram showing an example of the hardware configuration of a computer device that implements a server. [Figure 2B] Figure 2B is an explanatory diagram showing an example of the hardware configuration of a computer device that realizes a terminal device. [Figure 3] Figure 3 shows an example of the functions required to implement dialogue coaching. [Figure 4A]FIG. 4A is a diagram showing an example of the internal configuration of the dialogue output unit. [Figure 4B] FIG. 4B is a diagram showing an example of the internal configuration of the dialogue input unit. [Figure 5] FIG. 5 is a flowchart showing an example of the processing of the dialogue coach. [Figure 6] FIG. 6 is a diagram showing an example of the display of the learning UI. [Figure 7] FIG. 7 is a flowchart showing a specific example of the dialogue by the dialogue coach. [Figure 8] FIG. 8 is a diagram showing an example of the data of the learning scenario. [Figure 9] FIG. 9 is a diagram showing an example of the input prompt input to the generative AI.

Embodiments for Carrying Out the Invention

[0017] Hereinafter, with reference to the accompanying drawings, preferred embodiments of the dialogue-type teaching material device, dialogue-type teaching material program, and dialogue-type teaching material processing method according to this invention will be described in detail. <00\00096> In the embodiment, it is applied to dialogue-based learning that progresses learning through dialogue with a learner. The dialogue-type teaching material device of the embodiment is configured to be able to correctly determine the correctness of an answer that is not unique, such as in a free description format or a speech format other than a multiple-choice format with correct or incorrect answers. This enables continuous dialogue learning along the dialogue-based learning scenario.

[0019] (System Configuration of the Dialogue-Type Teaching Material Device) First, the system configuration of the dialogue-type teaching material device including the dialogue-type teaching material device according to the embodiment of this invention will be described.

[0020] Figure 1 is an explanatory diagram showing an example of the system configuration of an interactive teaching material device. In Figure 1, the interactive teaching material device 100 according to this embodiment of the invention consists of a server 110 and a plurality of terminal devices (interactive teaching material devices) 120. The server 110 is managed by the operator (administrator) of the interactive teaching material device 100. The server 110 can also be realized by a general-purpose computer device such as a personal computer (see Figure 2A). The server 110 stores a teaching material database and a subscriber database (both omitted from the illustration).

[0021] The terminal device 120 can be implemented by a computer device such as a tablet computer or tablet terminal (see Figure 2B). The terminal device 120 may be portable or not. The terminal device 120 may include, for example, a touchscreen consisting of a display and a touch panel (see Figure 2B). Touch panel operation can be performed using the learner's finger or a stylus 130.

[0022] The server 110 and each terminal device 120 are connected to each other via a network 140 such as the Internet, enabling them to communicate with one another. Each terminal device 120 can communicate wirelessly with radio repeaters installed on the network 140, for example, using a wireless LAN such as Wi-Fi (registered trademark), thereby achieving good portability for the terminal devices 120.

[0023] (Hardware configuration of Server 110) Figure 2A is an explanatory diagram showing an example of the hardware configuration of a computer device that implements a server. In Figure 2A, the computer device that implements the server 110 includes a CPU 211, memory 212, and a network interface 213. The various parts 211 to 213 of the computer device are connected by a bus 210. The CPU 211 functions as a control unit that oversees the overall control of the computer device that implements the server 110.

[0024] Memory 212 stores programs such as boot programs and data that make up various databases. Memory 212 also stores various databases, such as educational material databases and subscriber databases. These databases are stored in non-volatile storage media, which are among the various storage media that make up Memory 212, and whose contents are not erased even when the power is off.

[0025] Furthermore, memory 212 is used as the work area of ​​the CPU 211. Memory 212 can be implemented using, for example, ROM (Read-Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), and HD (Hard Disk).

[0026] The network interface 213 is connected to a network 140, such as the internet, and is connected to external devices such as terminal devices 120 via the network 140. The network interface 213 controls the interface between the network 140 and the internal workings of the computer device that makes up the server 110, and controls the input and output of data between the computer device that makes up the server 110 and the external devices.

[0027] (Hardware configuration of terminal device 120) Figure 2B is an explanatory diagram showing an example of the hardware configuration of a computer device that realizes a terminal device. In Figure 2B, the computer device that realizes the terminal device 120 includes a CPU 221, memory 222, network I / F 223, touchscreen 225, microphone 226, speaker 227, and camera 228. Furthermore, in the interactive teaching material device 100 according to this embodiment of the invention, each of the parts 221 to 228 of the computer device that realizes the terminal device 120 is connected by a bus 220.

[0028] The CPU 221 functions as a control unit that oversees the overall control of the terminal device 120. The memory 222 stores data such as the boot program. The memory 222 is also used as the work area for the CPU 221. The memory 222 can be implemented using, for example, flash memory.

[0029] The network interface 223 is connected to a network 140, such as the internet, via a Wi-Fi router or wired connection. The network interface 223 controls the input and output of data between the network 140 and external devices such as a server 110.

[0030] The touchscreen 225 comprises a display 225a and a touch panel 225b.

[0031] The display 225a can be implemented, for example, primarily by a liquid crystal display (LCD) or an electro-luminescence (OLED) display. The display 225a may be a color display or a monochrome (black and white) display.

[0032] The touch panel 225b is stacked on the display surface side of the display 225a and outputs a signal to the CPU 221 according to the operation position. For example, the touch panel 225b determines whether or not an input operation has been performed on the touch panel 225b, the position where the input operation was received on the touch panel 225b, the medium on which the input operation was performed, etc., and outputs a signal to the CPU 221 according to the received input operation.

[0033] Each time the terminal device 120 receives input information, it generates input data, generates transmission information that associates the identification information of the learning content at the time of input with the identification information of the terminal device 120, and sends the transmission information to the server 110 each time it is generated. When the server 110 receives the transmission information, it stores the identification information of the learning content and the identification information of the terminal device 120 contained in the transmission information in the memory 222.

[0034] The information to be transmitted may be stored in memory 222 for a certain period of time, such as until a certain amount of data is reached or until the learning of one piece of learning content is completed, and then transmitted to the server 110 when a certain amount of data is reached or when the learning of one character is completed.

[0035] Microphone 226 converts audio, such as a speaker's voice, which is input as analog data, from analog to digital and generates audio data in digital format. Microphone 226 is used, for example, when a learner pronounces something. Speaker 227 converts the digital audio data from digital to analog and outputs sound by energizing a coil in the speaker cone based on the analog audio data. Speaker 227 outputs, for example, a predetermined message. Speaker 227 may also output sounds such as a warning alarm.

[0036] Camera 228 generates image data through operation of the touch panel 225b. The generated image data can be stored in memory 222. Network I / F 223 is connected to a network 140 such as the internet and provides an interface for communication with server 110 via the network 140.

[0037] In the example configuration shown in Figure 2B, the terminal device 120 is shown as a hardware configuration equivalent to a tablet or smartphone. However, the terminal device 120 is not limited to this; it may also be a PC, in which case the touchscreen 225 shown in Figure 2B may be replaced with a display and mouse.

[0038] (An example of a teaching material database) The learning materials database stores learning content related to learners' studies. This learning content consists of, for example, "question" content and "answer" content. The "question" content includes, for example, video data, image data, and audio data related to the learning questions. The difficulty level of the questions can be varied by referencing data such as the learner's accuracy rate in the same subject area and the learner's attributes.

[0039] The "Answer" content holds metadata for the correct answer and includes information that allows for determination of whether it matches the learner's answer. For example, based on the learner's selection from multiple options displayed as images in the "Question" content as answers to the question, it can be determined whether the learner's answer to the question is correct or not.

[0040] Each "answer" content is stored in the learning materials database, associated with each "question" content. Each "answer" content is stored in the learning materials database, associated with one "question" content. In the learning materials database, at least one "answer" content is stored associated with each "question" content.

[0041] (An example of a subscriber database) The subscriber database stores information about the learning progress associated with each learner's identification information. This learning progress information represents the learning progress. This learning progress information can be implemented, for example, by information about the playback history of learning content. Alternatively, this learning progress information may be implemented, for example, by information about "question" content that has been displayed on the display 225a of the terminal device 120, or by information about "question" content for which an answer has been entered.

[0042] Information regarding the progress of learning may include, for example, information identifying the learning content, such as playback of the learning content, display of the learning content, and acceptance of the input of answers, as well as information regarding the date and time when the playback / display / acceptance of the input of answers for the learning content / the above determination was made (hereinafter referred to as "learning date and time" as appropriate).

[0043] In this way, by managing the learning progress for each learner, it is possible to provide appropriate learning tailored to the individual learner's progress. Furthermore, the subscriber database may store information about the learner themselves, such as the learner's name (member name), grade level, and date of birth, who is a subscriber to the learning service using the interactive learning material device 100. The subscriber database may also include information about the learner's guardian, such as the learner's parent.

[0044] (About dialogue coaching) Embodiments of the present invention include, as a learning material database, a dialogue coach in addition to the "problem" content and "answer" content described above. The dialogue coach has the function of presenting problems from the "problem" content and engaging in dialogue with the learner in the "answer" content, according to a predetermined learning scenario, as the learner progresses through the learning process. The dialogue is conducted using text and audio.

[0045] The dialogue coach initiates the conversation with the learner by asking questions, and displays the history of the conversation, for example, on a chat screen. The questions from the dialogue coach include explanations of problems (explanatory elements) and questions related to problems (question elements), and these explanations and questions are displayed on the chat screen. When the dialogue coach displays a question for the learner, it also displays a response field for the learner, and the learner enters their answer in the response field.

[0046] The dialogue coach displays one or a combination of multiple-choice, free-response, or spoken response formats as answer fields. The multiple-choice format supports closed questions where the answer must be unique, and the response can be displayed as, for example, Yes, Yes / No, or No.

[0047] In the free-response format, learners answer questions by typing their own words. The dialogue coach handles the text input by recognizing not only text data entered via keyboard but also image data input from handwritten input using a pen, etc., and converting it into text data. In the spoken format, learners answer questions by speaking their own words in response to the audio questions.

[0048] The dialogue coach converts the text data of the problem into speech and outputs it to the learner through a speaker. It also captures the audio data of the learner's response using a microphone and converts it into text data. The speech recognition and text recognition functions can utilize general-purpose technologies such as APIs (Application Programming Interfaces).

[0049] The dialogue coach performs a unique correct or incorrect determination for multiple-choice answers. For free-response and dialogue-based answers, the dialogue coach uses generative AI (Artificial Intelligence) to determine the correctness of the answer. Generative AI can utilize general-purpose technologies such as ChatGPT.

[0050] The dialogue coach uses a generative AI to evaluate responses in both free-response and dialogue formats, classifying them into three categories: correct, incorrect, and partially incorrect (almost correct). The generative AI learns the learning scenario and the attributes of the answer choices (correct, incorrect, partially incorrect) to determine the correctness of the learner's responses.

[0051] (Examples of functions for realizing dialogue coaching) Figure 3 shows an example of the functionality of the dialogue coach. The dialogue coach functionality described above can be realized, for example, by the CPU 221 of the terminal device 120 executing the dialogue coach program. However, it is not limited to this; the CPU 211 of the server 110 may perform the dialogue coach program processing, and the terminal device 120 may mediate the dialogue between the server 110 and the learner.

[0052] The terminal device 120 includes a learning user interface (UI) 301 for communication with the learner and the functions of a dialogue coach 302. In the example in Figure 3, the learning UI 301 outputs the content of the dialogue (questions and explanations) output by the dialogue coach 302 to the learner using the terminal device 120. The learning UI 301 inputs the content of the learner's speech (answers to questions) to the dialogue coach 302. As a UI for audio input and output, the learning UI 301 has a microphone 226 that acquires the learner's voice and a speaker 227 that outputs the voice to the learner.

[0053] Furthermore, the dialogue coach 302 exchanges image data and text data with the learner. For example, during a dialogue with the learner on the terminal device 120, the dialogue coach 302 can output image data of questions and answer choices to the learner, and can also input not only the data of the choices selected by the learner, but also text data of the answers.

[0054] The dialogue coach 302 includes a dialogue output unit 311, a scenario extraction unit 312, and a dialogue input unit 313. The dialogue output unit 311 converts the problem and explanation data output by the scenario extraction unit 312 into speech or text and outputs it to the learner. The dialogue input unit 313 outputs the converted speech or text data of the learner's answers to the problems and explanations to the scenario extraction unit 312.

[0055] The scenario extraction unit 312 engages in dialogue with the learner based on the learning scenario. The learning scenario includes multiple chapters. The scenario extraction unit 312 outputs data such as explanations that correspond to the multiple chapters of the learning scenario. By continuing the dialogue with the learner through explanations and answers for each of the multiple chapters, the system helps the learner deepen their understanding of the problem and guides them to the answer.

[0056] As will be explained in more detail later, the scenario extraction unit 312 determines the learner's level of understanding of the problem and, depending on that level of understanding, either moves to the next chapter or engages in dialogue with the learner within the chapter to help them understand the problem.

[0057] Furthermore, the dialogue coach 302 is connected to the learning management unit 320, which manages the learner's learning, and can send logs of the learning status, including progress, to the learning management unit 320. The learning management unit 320 stores and manages logs for each learner in the subscriber database, etc., as described above.

[0058] Figure 4A shows an example of the internal configuration of the dialogue output unit, and Figure 4B shows an example of the internal configuration of the dialogue input unit. The dialogue output unit 311 shown in Figure 4A includes a dialogue learning engine 401, a speech engine 402, a speech dictionary 403, and a learning scenario S as functions related to utterance output to the learner.

[0059] The dialogue learning engine 401 refers to the learning scenario S corresponding to the scenario extracted by the scenario extraction unit 312 and outputs dialogue data with the learner. In dialogue learning, for example, a chat screen is used.

[0060] The dialogue learning engine 401 outputs the text data of the dialogue to the learning UI 301. Based on the dialogue data output by the dialogue learning engine 401, the speech engine 402 generates speech data for the dialogue and outputs it to the learning UI 301. The speech dictionary 403 stores the utterances and intonation during the dialogue and is referenced by the speech engine 402 during speech processing.

[0061] The dialogue input unit 313 shown in Figure 4B includes a speech recognition engine 411, a handwriting recognition engine 412, a character data conversion unit 413, and a generation AI 421 as functions related to speech input from the learner.

[0062] The speech recognition engine 411 recognizes speech input via the learning UI 301. The handwriting recognition engine 412 recognizes handwritten characters input via the learning UI 301. The character data conversion unit 413 converts the speech output by the speech recognition engine 411 and the handwritten characters output by the handwriting recognition engine 412 into character data, and outputs the character data to the dialogue learning engine 401.

[0063] The dialogue learning engine 401 determines whether the input options for a question are correct or incorrect if they are unique. On the other hand, if the answer to the question is not unique, it uses the generation AI 421 to determine whether the dialogue text data input via the learning UI 301 to the text data conversion unit 413 is correct or incorrect. Furthermore, if the learner outputs an instruction to proceed to the next scenario via the learning UI 301 (including no question asked), the dialogue learning engine 401 instructs the scenario extraction unit 312 to move to the next scenario.

[0064] When the dialogue learning engine 401 uses the generating AI 421, it specifies the prompt P it has as a parameter, outputs the learning scenario S as information to the generating AI 421, and obtains the generating AI 421's answer judgment. The prompt P is, for example, a question that contains multiple correct answers.

[0065] The generating AI 421 determines the answer in three stages: correct, incorrect, and partially incorrect (almost correct) based on the prompt P and the learning scenario S. The determination result is output to the learner via the dialogue output unit 311 to the learning UI 301 described above.

[0066] The dialogue coach 302 is not limited to being located on the terminal device 120 or the server 110; it may also be located outside the server 110. The dialogue coach 302 may also be located in a cloud on a network. Furthermore, the speech recognition engine 411 and handwriting recognition engine 412 of the dialogue coach 302 may utilize a general-purpose API (Application Programming Interface).

[0067] (Example of how a dialogue coach handles this) Figure 5 is a flowchart showing an example of the dialogue coach's processing. The dialogue coach 302 performs the following processing, for example, by executing a program on the CPU 221 of the terminal device 120 shown in Figure 2B. Note that each functional unit shown in Figures 4A and 4B is denoted by a reference numeral corresponding to the processing.

[0068] (1) First, as shown in Figure 4A, the dialogue coach 302 obtains the character data and parameters to be displayed from the learning scenario S (step S501). The dialogue coach 302 holds a variable that manages the display position on the chat screen and displays the character data at the initial display position (for example, the first line at the very top). The initial display position can be arbitrarily specified by the learning management unit 320.

[0069] (2) Next, the dialogue coach 302 outputs the text data to the learning UI 301, and the dialogue coach 302 generates speech data via the speech engine 402 and outputs it to the learning UI 301 (step S502). At this time, the dialogue coach 302 updates a variable that manages the display position of the text data, for example, to display the text data on the second line. At this time, the dialogue coach 302 adds the speech dictionary 403, which defines the speech intonation and pronunciation that has been prepared in advance so that appropriate speech can be produced, to the speech engine 402.

[0070] (3) Next, the dialogue coach 302 plays text data and audio data related to the problem and explanation via the learning UI 301 (step S503) and waits for the learner's input (step S504: loop of No).

[0071] (4) After this, as shown in Figure 4B, if the learner inputs an answer to a question or explanation via the learning UI 301 (step S504: Yes), the dialogue coach 302 accepts this input.

[0072] (5) If the learner's input is voice data or handwritten character data, the dialogue coach 302 recognizes these using the speech recognition engine 411 and the handwriting recognition engine 412, and converts them into character data using the character data conversion unit 413 (step S505). If the learner's input is a single-choice answer, the speech recognition engine 411 and the handwriting recognition engine 412 do not perform recognition processing, and the answer is directly input to the dialogue learning engine 401.

[0073] (6) The dialogue coach 302 then uses the dialogue learning engine 401 to determine the learner's answer in three stages: correct, incorrect, or partially incorrect (almost correct) (step S506).

[0074] (7) Also, if there are parameters to use the generating AI421 in step S501, the "prompt P specification + learning scenario S" that it has as a parameter is passed to the generating AI421 as information, and the system waits for the generating AI421 to make a response determination (step S507).

[0075] (8) The dialogue coach 302 passes the result of the response evaluation to the scenario extraction unit 312 (step S508), and returns to the process of step S501.

[0076] (9) The dialogue coach 302 sends a log of the above processes to the learning management unit 320 as needed.

[0077] (Example of the learning UI display) Figure 6 shows an example of the learning UI display. The display example shown in Figure 6(a) is an example of a question from a middle school science subject. At the top of the display screen 600, the question explanation 601, "Figures A and B below show what happens when the surface of either granite or andesite is polished and observed at the same magnification to see how they are formed," is displayed, and below that, images 602 of A and B are displayed. In addition, question 603 is displayed, "(1) When the structure of a certain type of rock different from A and B was observed, this rock had a similar structure to A, but was darker in color than A. Choose the correct name of this rock from the following options A to D and answer with the letter."

[0078] The correct answers to the question are "volcanic rock," "rhyolite," "andesite," and "basalt." These belong to the igneous rock group. Although not shown in Figure 6, one of these four rocks ("volcanic rock," "rhyolite," "andesite," and "basalt") is displayed as an option from A to D, and the other three are different.

[0079] Furthermore, if the response is in a free-response or dialogue format and not unique, the dialogue coach 302 uses a generation AI to determine the answer in three stages: correct, incorrect, or partially incorrect (almost correct).

[0080] Learners can activate the dialogue coach 302 and display it on the screen through any operation. Alternatively, the dialogue coach 302 may be activated automatically. Upon activation of the dialogue coach 302, the chat screen 610 is displayed as an overlay on the display screen 600 shown in Figure 6. For example, learners can activate the dialogue coach 302 if they cannot answer immediately or if they want to answer while listening to an explanation.

[0081] The chat screen 610 displayed when the conversation coach 302 is launched displays new messages sequentially from top to bottom, just like a general-purpose chat screen. In the example in Figure 6, the conversation coach 302 displays, on the first line, an explanation of the problem 611, "When magma solidifies rapidly, it becomes volcanic rock, and when magma solidifies slowly, it becomes plutonic rock," in accordance with the learning scenario S, and on the second line, an explanation 612, "Because volcanic rock solidifies rapidly, the minerals in the rock cannot grow and become small." These displays correspond to a review (repetition) of content learned in the past.

[0082] Then, on the third line, question 613 is displayed: "Based on what we've seen so far, what kind of rock do you think A is?" Question 613 corresponds to the knowledge required to answer problem 603. The dialogue coach 302 can display the text on the chat screen 610 as an image and output it as audio.

[0083] A response field 620 is provided at the bottom of the chat screen 610. Learners enter their answers to the question 613 on the chat screen 610 into the response field 620. At the top of the response field 620 are a text input button 621, a voice recognition button 622, and a handwriting recognition button 623. Learners can input their answers using text, voice, or handwriting by selecting from these buttons.

[0084] For example, in the answer field 620 of Figure 6(a), the learner has selected the handwriting recognition button 623 and entered the handwritten characters "volcanic rock". Also, as shown in Figure 6(b), if the learner selects the voice recognition button 622, the system becomes ready to accept voice input, and "Voice Recognition in Progress" is displayed on the chat screen 610, allowing the learner to input by voice. After a certain period of time, as shown in Figure 6(c), the voice recognition result "volcanic rock" is displayed in the answer field 620.

[0085] At the bottom of the chat screen 610, there is a send button 640, which allows the learner to send (output to the conversation coach 302) the recognition results of handwriting recognition or speech recognition entered in the answer field 630. If there is no content to send in the answer field 630, the send button 640 will be displayed in an inconspicuous color such as gray and will be unavailable.

[0086] (Specific examples of dialogues by a dialogue coach) Figure 7 is a flowchart illustrating a specific example of a dialogue conducted by a dialogue coach. Based on the learning scenario S, the dialogue coach 302 branches the scenario into different routes according to the learner's level of understanding of the problem. Figure 7 shows the correct answer route A, the explanation route B, and the rescue route C, corresponding to the three-value judgment results of the answer to question 613, "Based on what we've seen so far, what kind of rock do you think A is?" in the chat screen 610 shown in Figure 6.

[0087] Route A is the scenario route when the answer to the question is correct, Route B is the scenario route when the answer is partially incorrect (almost correct), and Route C is the scenario route when the answer is incorrect. Figure 7 shows the dialogue in text, but it can also be done in combination with audio or as audio only.

[0088] To explain the chat-based dialogue state according to Figure 7, if the learner's answer to question 613, "Based on what we've seen so far, what kind of rock do you think A is?" is "The names of igneous rocks such as volcanic rock, rhyolite, andesite, and basalt (correct answer)", then dialogue coach 302 will determine that it is correct and, following the processing of correct answer route A, will respond to question 613 on chat screen 610 with response S701, "Excellent! That's correct!".

[0089] Next, dialogue coach 302 responds with explanation S702, "Now we know that A is volcanic rock. Let's take a closer look at volcanic rock." After this, the dialogue coach outputs other explanations and questions to the learner to help them understand the original problem 603, using the same process as in Figure 6, and the dialogue progresses based on the learner's responses.

[0090] Furthermore, when the learner's answer to question 613 is "igneous rock," the dialogue coach 302 determines that it is partially incorrect (almost correct). In this case, following the processing of explanation route B, the dialogue coach 302 responds to question 613 on chat screen 610 with explanation S711, "Almost! I'd like you to be a little more specific than just saying igneous rock," in order to further improve the learner's understanding and lead them to the correct answer. The dialogue coach 302 then responds with question S712, "What kind of rock do you think A is?"

[0091] Next, if the learner's answer is "the names of igneous rocks such as volcanic rock, ripple, andesite, and basalt (correct answer)," dialogue coach 302 will determine that it is correct and respond S713 "Excellent! That's correct!" following the processing of explanation route B. After this, dialogue coach 302 will move to correct answer route A and respond with explanation S702.

[0092] Furthermore, if the learner's answer to question 613 is incorrect (other than the names of igneous rocks such as volcanic rock, rhyolite, andesite, and basalt), the dialogue coach 302 will determine that it is incorrect and, following the processing of rescue route C, will respond to question 613 on the chat screen 610 with explanation S721: "Too bad! You should notice that the minerals in rock A are smaller than those in rock B."

[0093] Next, if the learner's answer is incorrect (other than the names of igneous rocks such as volcanic rock, rimonite, andesite, and basalt), dialogue coach 302 will determine that it is incorrect and respond with the correct answer S722: "Too bad! A has smaller minerals in the rock than B, so the correct answer is volcanic rock, which is formed when magma solidifies rapidly." After this, dialogue coach 302 will move to correct answer route A and respond with explanation S702. In this way, in rescue route C, if the learner continues to give incorrect answers even after multiple explanations, the correct answer will ultimately be conveyed to the learner.

[0094] (Example of training scenario data) Figure 8 shows an example of data for a learning scenario. Figure 8(a) shows an example of data for learning scenario S corresponding to the explanations in Figures 6 and 7, and is labeled with the corresponding symbols in the explanations in Figures 6 and 7.

[0095] The dialogue coach 302 searches for a match with the string responded to by the generating AI 421 and proceeds to process the line number after the colon. For example, if the response from the generating AI 421 to question (jump) 613 at line number 67, "Based on what we've seen so far, what kind of rock do you think A is?", is correct, the dialogue coach 302 then responds with response (msg) S701 at line number 69, "Excellent! That's correct!". If the response from the generating AI 421 to question 613 is incorrect, the dialogue coach 302 then responds with responses S711 at lines 194 and 195, "Almost! I'd like you to be a little more specific than just saying it's an igneous rock." The last string at line number 67, "json67.json", is the filename of the additional information (extended area).

[0096] Figure 8(b) shows an example of the detailed data of the "json67.json" file, including the information flag 801 "useGenAIRecog":true, which indicates whether to perform judgment using the generated AI421, and the content of the added prompt P. The content of prompt P is "prompt":"Use the condition that best applies from the following conditions to determine the answer to this question and output it with tags."..., followed by answer information for correct answer 1, incorrect answer 1 corresponding to a partially incorrect answer, incorrect answer 2 corresponding to an incorrect answer, and information on the judgment result output format.

[0097] The information for each correct answer 1 and incorrect answer 1 and 2, as well as the output format of the judgment result, can be found, for example: "\n###Judgment Method\n\n·If the answer is equivalent to the model answer (volcanic rock, rhyolite, andesite, basalt), it is "Correct 1"\n·If the answer is closest to an answer that deviates from the model answer (e.g., igneous rock), it is "Incorrect 1"\n·If any other answer does not fit the context, it is "Incorrect 2"\n###Output Format\nExample: <result> correct answer< / result> ￥n￥n <result> (Judgment result)< / result> It is "￥n"」.

[0098] As shown in Figure 8(b), if the information flag 801 is true, the above content is added to prompt P, indicating that the generated AI421 will be used to determine the answer to question 613 on line 67.

[0099] Figure 9 shows examples of input prompts that are input to the generating AI. Figure 9(a) is prompt 1 (system prompt), and Figure 9(b) shows an additional prompt 2 (user prompt) related to prompt 1.

[0100] Based on the detailed information of the scenario data S shown in Figure 8 (json67.json in Figure 8(b)), the dialogue coach 302 will use prompt 1 shown in Figure 9(a) as follows: "Use the condition that best applies from the following options to determine the answer to this question and output it with tags included." ###Judgment method If your answer is equivalent to the model answer (names of igneous rocks such as volcanic rock, rhyolite, andesite, and basalt), then it is marked as "Correct Answer 1". If your answer is equivalent to the model answer (igneous rock), it will be marked as "Incorrect 1". • Any other answer that does not fit the context will be marked as "Incorrect 2". ###Output format example: <result> correct answer< / result> ￥n￥n <result> (Judgment result)< / result> Input "" into the generation AI421.

[0101] Furthermore, prompt 2 in Figure 9(b) is an additional prompt 2 corresponding to question 613 (see Figure 6), and is used at the time when question 613 is asked to the learner. Enter "Based on what we've seen so far, what kind of rock do you think A is?" into the generating AI421.

[0102] Here, there may be multiple correct answers, such as correct answer 1, correct answer 2, correct answer 3, ... and similarly, there may be multiple incorrect answers. The incorrect answer with the highest ordinal number (incorrect answer 2 in the above example) is a completely incorrect answer. Dialogue coach 302 treats all others as "partially incorrect (almost correct)."

[0103] Then, as shown in Figure 9(c), the generated AI421 has a response pattern, " <result> (Judgment result)< / result> " is the response to dialogue coach 302. Dialogue coach 302 says, <result>The branching process for the above scenario is performed using the judgment result enclosed in tags.

[0104] In the process described above, the explanation has been simplified for convenience, but for example, there may be cases where the number of items included in the correct group is much larger. However, according to this embodiment, even when the number of items included in the correct group is large, the generation AI can be used to correctly determine whether the result is correct or incorrect.

[0105] In the embodiments described above, an example was described in which judgment is made in three stages: correct, partially incorrect, and incorrect. The present invention is not limited to this, and more detailed stages may be added between correct and incorrect. For example, it is possible to judge in four or more stages, such as completely correct, almost correct, partially correct, and incorrect, depending on the accuracy of the answer.

[0106] Furthermore, in the judgment by the generating AI, an evaluation score may be assigned individually to each correct element included in the answer, and the determination of whether the answer is correct, partially incorrect, or incorrect may be made based on the sum of the evaluation scores of each correct element. For example, it is also possible to set a point value for each element included in the model answer and make a judgment based on the sum of the points of the elements included in the answer.

[0107] As described above, the interactive teaching material device of the embodiment is an interactive teaching material device that conducts learning through dialogue with the learner, and includes a control unit that performs the following: a process of asking the learner a predetermined question; a process of receiving an answer to the question from the learner; and a process of using a generating AI to determine whether the learner's answer is correct, incorrect, or partially incorrect.

[0108] This makes it easy to determine a learner's answer even if it falls into multiple categories, such as three levels of correct, incorrect, and partially incorrect. Since partially incorrect answers involve multiple responses, existing technologies require preparing a large number of anticipated answers in advance, resulting in high man-hours and costs. However, using generative AI allows for easy, low-cost, and accurate determination of correctness. Furthermore, the dialogue built with the learner allows for an understanding of the question and the underlying problem. Learners can think deeply about the problem, solve problems they couldn't solve before, and understand the process of solving them. This also promotes personalized learning using devices such as tablets.

[0109] Furthermore, in the interactive teaching material device of this embodiment, the response can be one of the following: selection of a predetermined unique option, a free-form text input, or a dialogue format using arbitrary speech. The control unit uses a generation AI to make judgments on the responses in the free-form text and dialogue formats.

[0110] This allows learners to respond in various formats, increasing the degree of freedom in their answers. Furthermore, using generative AI enables accurate evaluation of both free-response and dialogue-based answers.

[0111] Furthermore, in the interactive teaching material device of the embodiment, the control unit may, in judgment using the generating AI, individually assign evaluation points to each correct element included in the answer, and make a judgment in multiple stages, including correct, partially incorrect, and incorrect, based on the sum of the evaluation points for each correct element.

[0112] This makes it possible to easily and appropriately determine multiple levels of accuracy, including correct, partially incorrect, and incorrect answers, based on the total score.

[0113] Furthermore, in the interactive teaching material device of this embodiment, the control unit branches into different scenarios based on the level of understanding of the question, according to the judgment result.

[0114] This will enable the creation of dialogues tailored to each learner's level of understanding of the questions.

[0115] Furthermore, the interactive teaching material device of the embodiment includes, as information to be input to the generating AI, a control unit containing information on a method for determining a question, an output format after determination, and a set of pre-prepared answers to the question included in the scenario.

[0116] This will enable the accurate detection of answers generated by AI, especially those that are partially incorrect.

[0117] Furthermore, the interactive teaching material device of the embodiment has a control unit which includes a handwriting recognition engine that recognizes characters written in a free-form writing style and a speech recognition engine that recognizes speech spoken in a dialogue style.

[0118] This allows learners to respond using voice or handwritten text as they see fit, making it easier to conduct the conversation.

[0119] Furthermore, the interactive teaching material device of the embodiment includes a control unit that provides explanations to guide the learner to the correct answer, according to the learner's level of understanding of the question.

[0120] This allows learners to deepen their understanding of questions and problems by listening to explanations tailored to their level of comprehension.

[0121] The processing method using the interactive teaching materials described in this embodiment can be implemented by executing a pre-prepared program on a computer such as a personal computer or workstation. This program is recorded on a computer-readable recording medium such as a hard disk, flexible disk, CD-ROM, MO, or DVD, and is executed by being read from the recording medium by the computer. This program may also be transmitted via a network such as the Internet. [Industrial applicability]

[0122] As described above, the interactive teaching material device, interactive teaching material program, and interactive teaching material processing method according to this invention are useful for interactive teaching material devices, interactive teaching material programs, and interactive teaching material processing methods that support learning, and are particularly suitable for interactive teaching material devices, interactive teaching material programs, and interactive teaching material processing methods that enable learners to understand problems and their solutions through dialogue with them. [Explanation of Symbols]

[0123] 100 Interactive Teaching Materials Devices 110 Servers 120 Terminal devices 301 Learning UI 302 Dialogue Coach 311 Dialogue Output Unit 312 Scenario Extraction Unit 313 Dialogue Input Section 401 Dialogue Learning Engine 402 Speech Engine 411 Speech Recognition Engine 412 Handwriting Recognition Engine 413 Character Data Conversion Unit 421 Generation AI 600 display screen 610 Chat screen P prompt S Learning Scenario< / result>

Claims

1. In an interactive teaching material device that facilitates learning through dialogue with learners, The process involves asking learners predetermined questions, The process of receiving answers from learners to the aforementioned questions, The process involves using a generative AI to determine whether the learner's answer is correct, incorrect, or partially incorrect, and An interactive teaching material device characterized by having a control unit that performs the following.

2. The aforementioned response can be one of the following: selection of a predetermined unique option, free-form text input, or dialogue using arbitrary utterances. The interactive teaching material device according to claim 1, characterized in that the control unit makes a determination using the generating AI for the answers in the free-response format and the dialogue format.

3. The interactive teaching material device according to claim 2, characterized in that the control unit, in the judgment using the generating AI, individually assigns evaluation points to each correct element included in the answer, and makes a judgment in multiple stages, including correct, partially incorrect, and incorrect, based on the sum of the evaluation points of each correct element.

4. The interactive teaching material device according to claim 1, characterized in that the control unit branches to a scenario corresponding to the level of understanding of the question based on the result of the determination.

5. The interactive teaching material device according to claim 4, characterized in that the control unit includes, as information to be input to the generating AI, information on a method for determining the question, an output format after determination, and a plurality of pre-prepared answers to the question included in the scenario.

6. The interactive teaching material device according to claim 2, characterized in that the control unit has a handwriting recognition engine that recognizes characters written in the free-form writing format and a speech recognition engine that recognizes speech spoken in the dialogue format.

7. The interactive teaching material device according to claim 1, characterized in that the control unit includes a process for providing explanations to guide the learner to the correct answer according to the learner's level of understanding of the question.

8. In interactive learning materials programs where learning takes place through dialogue with learners, The process involves asking learners predetermined questions, The process of receiving answers from learners to the aforementioned questions, The process involves using a generative AI to determine whether the learner's answer is correct, incorrect, or partially incorrect, and An interactive educational program characterized by having a computer execute it.

9. In an interactive material processing method that involves learning through dialogue with learners, The process involves asking learners predetermined questions, The process of receiving answers from learners to the aforementioned questions, The process involves using a generative AI to determine whether the learner's answer is correct, incorrect, or partially incorrect, and An interactive teaching material processing method characterized by a computer executing the following.

Images