system

Abstract

The system according to the present embodiment includes a reception unit, an analysis unit, a proposal unit, and a selection unit. The reception unit receives an input from a user. The analysis unit analyzes a context based on the input received by the reception unit. The proposal unit proposes a proper noun based on the context analyzed by the analysis unit. The selection unit allows the user to select the proper noun proposed by the proposal unit.

Description

Technical Field

[0001] The technology disclosed herein relates to a system. Background Technology

[0002] Patent Document 1 discloses a personalized chatbot control method executed by at least one processor, the method comprising: receiving user speech; adding the user speech to a prompt containing instructions related to a chatbot role; encoding the prompt; and inputting the encoded prompt into a language model to generate chatbot speech in response to the user speech.

[0003] Patent document 1: Japanese Patent Application Publication No. 2022-180282.

[0004] In the existing technology, there is a problem that when one cannot recall proper nouns in an instant messaging application, there are not enough means to smoothly assist in input. Summary of the Invention

[0005] The system described in this embodiment includes a receiving unit, a parsing unit, a proposal unit, and a selection unit. The receiving unit receives user input. The parsing unit parses the context based on the input received by the receiving unit. The proposal unit proposes proper nouns based on the context parsed by the parsing unit. The selection unit allows the user to select proper nouns proposed by the proposal unit. Attached Figure Description

[0006] Figure 1 This is a conceptual diagram illustrating an example of the configuration of a data processing system according to the first embodiment.

[0007] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

[0008] Figure 3 This is a conceptual diagram illustrating an example of the data processing system configuration in the second embodiment.

[0009] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

[0010] Figure 5 This is a conceptual diagram illustrating an example of the data processing system configuration in the third embodiment.

[0011] Figure 6 This is a conceptual diagram illustrating an example of the main functions of the data processing device and head-mounted terminal according to the third embodiment.

[0012] Figure 7 This is a conceptual diagram illustrating an example of the data processing system configuration in the fourth embodiment.

[0013] Figure 8 This is a conceptual diagram illustrating an example of the functions of the main parts of the data processing device and robot according to the fourth embodiment.

[0014] Figure 9 It represents an emotion graph that maps multiple emotions.

[0015] Figure 10 It represents an emotion graph that maps multiple emotions.

[0016] Explanation of reference numerals in the attached figures Data processing systems 10, 210, 310, and 410 12 Data processing devices 14 Smart devices 214 Smart Glasses 314 Head-mounted terminal 414 Robot. Detailed Implementation

[0017] Hereinafter, an example of an implementation of the system involved in this disclosure will be described with reference to the accompanying drawings.

[0018] First, let's explain the terms used in the following description.

[0019] In the following embodiments, the processor (hereinafter referred to as "processor") can be a single computing device or a combination of multiple computing devices. Furthermore, a processor can be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), or TPU (Tensor Processing Unit), etc.

[0020] In the following implementation, the labeled RAM (Random Access Memory) is a memory that temporarily stores information and is used by the processor as working memory.

[0021] In the following embodiments, the labeled memory is one or more non-volatile storage devices used to store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disk (e.g., hard disk) or magnetic tape, etc.

[0022] In the following implementation, the labeled Communication I / F (Interface) is an interface that includes a communication processor and an antenna, etc. The Communication I / F is responsible for communication between multiple computers. Examples of communication standards applicable to the Communication I / F include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0023] In the following implementation, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects more than three items, the same approach as "A and / or B" applies.

[0024] [First Implementation] Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0025] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. An example of the data processing device 12 is a server.

[0026] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 includes a computer 36, a receiver 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. In addition, the receiver 38, output device 40, and camera 42 are also connected to the bus 52.

[0028] The receiving device 38 includes a touchscreen 38A and a microphone 38B, etc., for receiving user input. The touchscreen 38A receives user input generated by contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input generated by sound by detecting the user's voice. The control unit 46A sends data representing user input received via the touchscreen 38A and microphone 38B to the data processing unit 12. In the data processing unit 12, a specific processing unit 290 (see...) Figure 2 Get the data that represents user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, etc., and presents data to the user by outputting data in a user-perceptible form (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

[0031] Figure 2 An example of the main functions of the data processing device 12 and the smart device 14 is shown.

[0032] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0033] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0034] In the smart device 14, specific processing is performed by the processor 46. A specific processing program 60 is stored in the memory 50. The specific processing program 60 is used in conjunction with the data processing system 10. The processor 46 reads the specific processing program 60 from the memory 50 and executes the read specific processing program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific processing program 60 executed on the RAM 48. Furthermore, the smart device 14 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0035] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing performed by the data processing system 10 of the first embodiment will be described.

[0036] (Example) The system described in this invention is designed to assist users in recalling specific proper nouns when composing conversations using instant messaging applications. When a user encounters a situation where they cannot recall a proper noun, such as "The lead actress of that movie is…", the AI ​​analyzes the context and suggests suitable proper nouns. The user can then select the suggested proper noun to reflect in the conversation. For example, when a user cannot recall a proper noun while composing a conversation in an instant messaging application, they can invoke AI input assistance. The AI ​​then analyzes the conversation context and suggests suitable proper nouns. For example, for a conversation like "The lead actress of that movie is…", the AI ​​might suggest the movie title or the actress's name. The user can then select the suggested proper noun to reflect in the conversation. This feature allows users to continue communicating smoothly even if they cannot recall proper nouns. Furthermore, because the AI ​​analyzes the context and suggests suitable proper nouns, users can accurately reflect information in the conversation. For example, when a user cannot recall a proper noun, a "processing unit" is needed to invoke AI input assistance. Next, a "parse department" is needed to analyze the dialogue context, and a "proposal department" is needed to propose appropriate proper nouns. Finally, a "selection department" is needed for the user to choose the proposed proper noun. These four main units—the receiving department, parsing department, proposal department, and selection department—are essential and interconnected. The receiving department receives user input, the parsing department analyzes the context, the proposal department proposes proper nouns, and the selection department receives user selections. As sub-units, for example, the parsing department may include a reference department to refer to past dialogue history or a general database; the proposal department may include a candidate suggestion department to suggest multiple candidate options; and the selection department may include a reflection department to reflect the user's choice. Thus, when writing dialogue in an instant messaging application, if a specific proper noun cannot be recalled, AI input assistance can be invoked and reflected in the dialogue. Specifically, when a user inputs text in an instant messaging application, the receiving department receives text data from the input interface (e.g., a UTF-8 encoded string, a context window with a maximum of 512 tokens). This receiving department detects patterns in user input that suggest missing proper nouns (e.g., "〇〇's leading actress is...", "〇〇's name is...") using regular expressions or triggered phrase dictionaries. When detected, the input data is transmitted to the parsing department. This parsing department refers to the input text, past dialogue history (e.g., up to 100 chronologically saved log entries per user in JSON format), and external knowledge bases (e.g., movie databases, character dictionaries, place name dictionaries, etc.) and extracts contextual features (e.g., topic category, most recent subject / predicate, and location of missing proper nouns) using natural language processing models (e.g., large-scale language models based on Transformer, pre-trained BERT or LSTM, etc.).The input examples for the AI ​​model include text such as "The lead actress of that movie is...", past dialogue history (e.g., "The movie I watched yesterday was 〇〇"), and a movie database (e.g., "The lead actress of movie 〇〇 is △△"). The AI ​​model outputs a candidate list of missing proper nouns from the input context (e.g., a list of names with fractions, in probability distribution form, up to 5 items). Output examples include "Candidates for lead actress: △△ (0.85), □□ (0.10), ×× (0.05)", etc. This proposal department receives the AI ​​model output and displays the candidate list on the user interface through the candidate suggestion department. The user selects a candidate through the selection department, and the response department automatically inserts the selection result into the dialogue text. The selection department detects the user's selection operation (e.g., click, tap, etc.) and records the selection log in the history database. This series of processes differs from traditional human memory retrieval or web page search; it is an unconventional algorithm based on AI that comprehensively calculates contextual features and knowledge base information in a high-dimensional vector space and scores the proper noun candidates. The AI ​​model is trained using a cross-entropy loss function and supervised learning datasets (such as matching movie titles with lead actresses, and matching conversational context with correct proper nouns). The technical effect is that even if the user cannot recall the proper noun, the system can still achieve uninterrupted smooth communication through the AI's high-speed and high-precision candidate suggestions. Furthermore, the improved accuracy of proper noun suggestions significantly reduces the hassle of inputting and retrieving incorrect information. Applicable fields include general instant messaging applications, business chat, customer support, educational dialogue systems, and medical case report assistance, among other text input assistance scenarios.

[0037] The instant messaging application input assistance system described in this embodiment includes a receiving unit, a parsing unit, a proposal unit, and a selection unit. The receiving unit receives user input. User input includes, for example, text input and voice input, but is not limited to these examples. For example, when a user cannot recall proper nouns while composing a conversation in the instant messaging application, the receiving unit invokes AI input assistance. The parsing unit parses the context based on the input received by the receiving unit. Context parsing includes, for example, the contextual relationships of the conversation and related topics, but is not limited to these examples. The parsing unit performs context parsing by referring to past conversation history or a general database, for example. The proposal unit proposes proper nouns based on the context parsed by the parsing unit. Proposal suggestions include, for example, names of people, places, and products, but are not limited to these examples. The proposal unit provides multiple candidate options, for example. The selection unit allows the user to select a proper noun proposed by the proposal unit. The selection unit reflects the user's selection, for example. Thus, even if the user cannot recall proper nouns during the conversation, they can continue communicating smoothly. Some or all of the above processing in the parsing unit can be implemented using AI, or AI may not be used. For example, the parsing unit can input user input into the AI, which then performs context parsing. Some or all of the above processing in the proposal unit can be implemented by AI, or AI may not be used. For example, the proposal unit can input the context parsed by the parsing unit into the AI, which then performs proper noun proposals. Some or all of the above processing in the selection unit can be implemented by AI, or AI may not be used. For example, the selection unit can input the proper nouns proposed by the proposal unit into the AI, which then performs user selections. Thus, the instant messaging application input assistance system according to this embodiment can accept user input, parse context, propose proper nouns, and reflect user selections. Specifically, this input assistance system receives text data (e.g., UTF-8 encoded strings, maximum 512 token context window) or voice data (e.g., 16kHz sampled PCM waveform, maximum 30 seconds) from the user input interface (e.g., software keyboard, voice recognition module) via the receiving unit. The receiving unit uses regular expressions or a triggered phrase dictionary (e.g., "The leading actress of 〇〇 is...", "The name of 〇〇 is...", etc.) to detect proper noun missing patterns in the input data, and when detected, transmits the input data to the parsing unit. This parsing unit takes the input text or speech and feeds it into the natural language processing module (e.g., a large language model based on Transformer, pre-trained BERT, LSTM, etc.) to extract contextual features (e.g., topic category, recent subject / verb, missing proper noun positions). Examples of input to the AI ​​model include text such as "The lead actress in that movie is...", past conversation history (e.g., "The movie I watched yesterday was 〇〇"), and a movie database (e.g., "The lead actress in movie 〇〇 is △△"). The AI ​​model outputs a candidate list of missing proper nouns from the input context (e.g., a list of names with fractions, in probability distribution form, up to 5 items).Output examples include "Leading actress candidates: △△ (0.85), □□ (0.10), ×× (0.05)", etc. The proposal department receives the AI ​​model output and displays the candidate list on the user interface through the candidate prompting department. Users select candidates through the selection department, and the response department automatically inserts the selection result into the dialogue text. The selection department detects the user's selection actions (e.g., clicks, taps, etc.) and records the selection log in the historical database. This series of processes differs from traditional human memory retrieval or web page search; it is an unconventional algorithm based on AI that comprehensively calculates contextual features and knowledge base information in a high-dimensional vector space and scores proper noun candidates. The AI ​​model is trained using a cross-entropy loss function and supervised learning datasets (e.g., matching movie titles with leading actresses, matching dialogue context with correct proper nouns, etc.). The technical effect is that even if users cannot recall proper nouns, the system can achieve uninterrupted smooth communication through high-speed and high-precision AI candidate prompts. Furthermore, the improved accuracy of proper noun proposals significantly reduces the hassle of inputting and retrieving incorrect information. It is applicable to various text input assistance scenarios, including general instant messaging applications, business chat, customer support, educational dialogue systems, and medical case reporting assistance.

[0038] The parsing unit may include a reference unit that references past dialogue history or a general database. For example, the reference unit references past dialogue history. Past dialogue history includes, but is not limited to, logs of past user conversations. The reference unit may also reference a general database. General databases include, but are not limited to, dictionary databases, knowledge bases, etc. Thus, by referencing past dialogue history or a general database, the accuracy of context parsing can be improved. Some or all of the above processing in the reference unit can be implemented using AI, or it may not use AI. For example, the reference unit can input past dialogue history or a general database into the AI, which will then perform the reference processing. Specifically, this parsing unit obtains up to 100 chronologically saved speech logs (JSON format, each speech with a timestamp, content, topic tags, etc.) and external knowledge bases (such as structured data like movie databases, biographical dictionaries, place name dictionaries, and product catalogs) for each user through the reference unit. This reference unit performs similarity calculations (such as cosine similarity, Euclidean distance, etc.) between the input contextual text (e.g., "The lead actress in that movie is...") and past dialogue history in a high-dimensional vector space, extracting historical records with high relevance. Furthermore, relevant proper noun entries can be retrieved from the knowledge base based on keywords or topic categories in the input context. When using AI, this reference unit inputs the input text, historical vectors, and knowledge base entities as multiple input tensors (e.g., 512 context tokens, 100 historical entries × 256-dimensional vectors, and 1000 knowledge base entities × 128-dimensional) into a large-scale language model based on Transformer or a pre-trained BERT, and calculates importance scores through the model's internal attention mechanism. Examples of AI model outputs include lists with scores such as "Relevant History: Speech ID123 (0.92), Speech ID87 (0.75)" and "Knowledge Base Candidates: Person A (0.88), Person B (0.65)". These outputs serve as inputs for subsequent context parsing or proper noun candidate generation. The AI ​​model is trained using historical data paired with correct proper nouns and knowledge base entity relevance annotation data, and weights are optimized using cross-entropy loss or triplet loss. This reference system differs from traditional human memory retrieval or simple keyword retrieval. It achieves unconventional, high-speed, and high-precision information extraction through similarity calculations and attention weighting in a high-dimensional vector space. The technical benefits include significantly improved contextual analysis accuracy by comprehensively referencing past dialogue history and knowledge bases through AI, resulting in a marked increase in the accuracy of locating and proposing the user's required proper nouns. Applicable areas include general instant messaging applications, business chat, automatic FAQ response, medical record assistance, and educational dialogue systems, making it suitable for various text input assistance scenarios where historical or knowledge base references are important.

[0039] The proposal department may include a candidate suggestion department that provides multiple options. The candidate suggestion department may, for example, provide multiple options. These multiple options may include, but are not limited to, proper noun candidates. The candidate suggestion department may, for example, provide the user with multiple proper noun candidates. Thus, by providing multiple options, the user has the option to choose. Some or all of the above processing in the candidate suggestion department may be implemented using AI, or it may not use AI. For example, the candidate suggestion department may input proper noun candidates into AI, which will then perform the candidate suggestion processing. Specifically, based on the contextual features, relevant history, and knowledge base candidates received by the parsing department, the proposal department generates a list of proper noun candidates (e.g., up to 5 items, each with a score or relevance label). The candidate suggestion department displays the candidates on the user interface in score order or by category (e.g., people, places, products, etc.). When using AI, the candidate suggestion unit inputs context vectors, history vectors, and knowledge base candidate vectors as multi-input tensors (e.g., 512-token context, 5-item × 128-dimensional candidate vectors) into a large-scale language model or classifier based on Transformer, outputting relevance scores or selection probability distributions for each candidate option. An example of the AI ​​model's output is a list with scores such as "Candidate A (0.85), Candidate B (0.10), Candidate C (0.05)". Based on these scores, the candidate suggestion unit prioritizes displaying the most relevant candidates, optimizing the visibility and operability of the selections. Furthermore, the candidate suggestion unit can add supplementary information to each candidate option (e.g., person's occupation, location, product name description, etc.). The AI ​​model is trained using data on matching correct proper nouns with the candidate list and ranking learning from user selection history (e.g., pairwise loss, listwise loss, etc.). This candidate suggestion unit differs from traditional simple list displays or static candidate suggestions; it achieves dynamic and high-precision candidate suggestions through AI's comprehensive scoring of context, history, and knowledge base. The technology allows users to easily select the optimal proper noun from a variety of options, reducing the risk of misselection or insufficient information. Applicable areas include general instant messaging applications, automated FAQ responses, customer support, medical record input assistance, and educational dialogue systems, suitable for various text input assistance scenarios where multiple candidate suggestions are helpful.

[0040] The selection section may include a reflection section that reflects user selections. For example, the reflection section reflects user choices. User choices may include, but are not limited to, the selection of proper nouns. The reflection section, for example, reflects the user-selected proper nouns into the dialogue. Thus, by reflecting user choices, appropriate proper nouns can be reflected into the dialogue. Some or all of the above processing in the reflection section can be implemented using AI, or it may not use AI. For example, the reflection section can input the user's selection into AI, which will then perform the reflection processing. Specifically, this selection section receives proper nouns selected by the user from a candidate list (e.g., click or tap selection operation, with option ID or score), and the reflection section automatically inserts the selection result into the missing position in the dialogue text. The reflection section also records the selection operation type (e.g., direct selection, voice command, shortcut key, etc.) and the context of the selection (e.g., recent speech content, selection time, etc.), and saves the selection log (e.g., option ID, selection time, user ID, context information, etc.) to a historical database. When using AI, the response unit inputs user-selected content and contextual information as input tensors (e.g., option ID, context vector, history vector, etc.) to the AI ​​model, outputting the optimal response method (e.g., automatic insertion position, whether to add explanations, insertion format, etc.). Examples of AI model outputs include "Insertion position: end of sentence, supplementary explanation: yes," and "Insertion format: bold." Based on these outputs, the response unit automatically inserts proper nouns into the user interface and adds supplementary information or emphasis when necessary. The AI ​​model is trained using paired data of selection history and response results, user satisfaction feedback, etc., applying reinforcement learning or supervised learning. This response unit differs from traditional simple text insertion or manual editing; through AI's comprehensive processing of context, history, and selected content, it achieves the optimal response that matches the user's intent and the situation. The technical effect is that users can instantly and accurately reflect proper nouns into the dialogue after making a selection, significantly reducing editing errors and operational burden. Applicable areas include general instant messaging applications, business chat, medical record input, educational dialogue systems, and automatic FAQ responses, suitable for various text input assistance scenarios where selection and response are important.

[0041] The processing unit can infer the user's emotions and adjust the timing of input assistance activation based on the inferred emotions. For example, the processing unit can infer the user's emotions, such as anxiety, relaxation, confusion, etc., but is not limited to these examples. For example, if the user is anxious, the AI ​​can immediately activate input assistance. If the user is relaxed, the processing unit can activate input assistance with a slight delay. If the user is confused, the processing unit can activate input assistance at an appropriate time. Thus, input assistance can be activated at the appropriate time based on the user's emotions. Emotion inference can be achieved, for example, through an emotion engine or generative AI. Generative AI can be text-generating AI (such as LLM) or multimodal generative AI, but is not limited to these examples. Some or all of the above processing in the processing unit can be implemented by AI, or AI can be omitted. For example, the processing unit can input the user's emotions into the AI, which will then perform emotion inference. Specifically, this processing department extracts multi-dimensional feature vectors from text data (e.g., UTF-8 encoded strings, maximum 512 tokens), voice data (e.g., 16kHz sampled PCM waveform, maximum 30 seconds), and input operation logs (e.g., key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (e.g., software keyboards, speech recognition modules). This processing department inputs these features into AI models for sentiment inference (e.g., BERT-based sentiment classifiers, multimodal Transformers, etc.). Examples of inputs to the AI ​​model include: "Input text: 'The leading actress in that movie is…'", "Input speed: 1.2 words / second", "Input interruption: 3 times", "Voice tone: high", etc. The AI ​​model outputs sentiment labels (e.g., anxiety, relaxation, confusion) and confidence scores (e.g., anxiety 0.78, relaxation 0.12, confusion 0.10) from these inputs. Examples of outputs include: "Emotion: Anxiety (0.78)" or "Emotion: Confusion (0.65)", etc. Based on the AI ​​model output, this reception department executes input assistance call timing decision algorithms (such as threshold judgment, priority scheduler, etc.). For example, it dynamically controls the call timing by immediately triggering an "anxiety" judgment, delaying it by 3 seconds for a "relaxation" judgment, and only calling it after detecting a stop in user input when a "confusion" judgment is detected. The AI ​​model is trained using sentiment-labeled input datasets (such as user input paired with sentiment self-report, voice / text / operation log combinations), applying cross-entropy loss function or multi-task learning. Subsequent processing uses sentiment inference results as the trigger for calling the input assistance module, optimizing the data transmission timing between the parsing and proposal departments. Unlike traditional simple timer or fixed rule calls, this reception department achieves unconventional timing control through AI multi-dimensional feature analysis and sentiment inference. The technical effect is that this system can provide input assistance timing based on the user's psychological state, thereby improving user experience, reducing input pressure, and minimizing the risk of session interruption.Applicable fields include general instant messaging applications, business chat, medical case input assistance in medical scenarios, educational dialogue systems, customer support, etc., and it is suitable for a variety of scenarios that require input assistance based on user emotions.

[0042] The processing department can analyze a user's past input history and select the optimal processing method. For example, the processing department analyzes the user's past input history. Past input history includes, but is not limited to, the user's past input logs. For example, the processing department prioritizes input methods frequently used by the user in the past. For example, the processing department detects specific patterns from the user's past input history and selects the optimal processing method. For example, the processing department proposes the most efficient processing method based on the user's past input history. Thus, by providing the optimal processing method based on the user's past input history, efficient input assistance is achieved. Some or all of the above processing in the processing department can be implemented using AI, or it can be done without AI. For example, the processing department can input the user's past input history into AI, which will then perform historical analysis. Specifically, this processing department retrieves the input history data (e.g., text input, voice input, image attachments, input device type, input time, input speed, number of corrections, etc., in JSON structure, up to 1000 records) for each user from the database, stored in chronological order. This service department extracts these historical data into feature vectors (e.g., input method category, input frequency, input success rate, input time period distribution, etc.), and inputs them into AI models (e.g., LSTM-based temporal pattern recognition model, Transformer-based historical analysis model, etc.). Examples of inputs to the AI ​​model include "Past 30 input methods: 25 text, 5 voice," "Input success rate: text 0.98, voice 0.85," and "Input time period: mostly at night." The AI ​​model outputs the optimal service method (e.g., prioritizing text input, prioritizing voice input, and combining image input) and recommendation scores (e.g., text 0.92, voice 0.08) from these inputs. Output examples include "Recommended service method: text (0.92)" or "Recommended service method: voice (0.75)," etc. Based on the AI ​​model's output, this service department automatically selects or prioritizes the displayed recommended service method on the user interface, facilitating efficient user input. The AI ​​model is trained using historical data paired with user satisfaction and input success rate data, applying cross-entropy loss function or reinforcement learning. The subsequent processing involves reflecting the recommended acceptance method in the initial settings of the acceptance module or triggering input assistance within the acceptance department. Unlike traditional methods that rely on simple historical references or static acceptance method selection, this acceptance department achieves efficient input through AI-powered multi-dimensional historical analysis and dynamic acceptance method optimization. The technical benefits include the system automatically proposing acceptance methods that align with user input preferences and efficiency, reducing input errors, increasing input speed, and optimizing user experience. Applicable areas include general instant messaging applications, business chat, medical record input assistance, educational dialogue systems, customer support, and more, suitable for various scenarios requiring optimal input acceptance for each user.

[0043] The processing department can automatically begin processing based on specific keywords or phrases triggered by the dialogue content. For example, the processing department can automatically begin processing based on specific keywords or phrases triggered by the dialogue content. Specific keywords or phrases include, but are not limited to, examples such as "the lead actress of ○ is...", "the location of ○ is...", and "the name of ○ is...". For example, the processing department can automatically begin processing when the user inputs "the lead actress of ○ is...", "the location of ○ is...", or "the name of ○ is...". Thus, by automatically starting processing based on specific keywords or phrases, the user's workload can be reduced. Some or all of the above processing in the processing department can be implemented using AI, or AI can be omitted. For example, the processing department can input the dialogue content into AI, which will then perform keyword or phrase detection. Specifically, this processing department monitors the text data received by the user input interface in real time (e.g., UTF-8 encoded strings, maximum 512 tokens), and uses regular expression pattern matching or triggers a phrase dictionary (e.g., a missing proper noun pattern of over 1000 entries) for keyword detection. In addition, the input text is fed into an AI model (e.g., a BERT-based context classifier, a Transformer-based trigger detection model, etc.) to accurately determine the location of missing proper nouns or areas requiring input assistance in the context. Examples of AI model inputs include "Input text: 'The lead actress in that movie is…'" and "Recent statement: 'The movie I watched yesterday was 〇〇'". The AI ​​model outputs trigger detection labels (e.g., proper noun missing trigger, ordinary statement, interrogative sentence, etc.) and detection confidence scores (e.g., trigger 0.95, ordinary 0.05) from the input context. An example output is "Trigger detection: proper noun missing (0.95)". Based on the AI ​​model output or pattern matching results, the processing department automatically activates the input assistance processing module and begins transmitting data to the parsing and proposal departments. The AI ​​model is trained using conversation datasets with trigger phrases or data with proper noun missing labels, applying cross-entropy loss or data augmentation methods. Subsequent processing uses the trigger detection results as the starting condition for the input assistance workflow, minimizing user operations and achieving automatic assistance. Unlike traditional simple keyword detection or static rules, this acceptance department uses AI contextual understanding and dynamic trigger judgment to achieve unconventional acceptance initiation control. The technical effect is that this system can automatically accept applications based on user input intent and context, significantly reducing the operation and latency of input assistance calls. Applicable areas include general instant messaging applications, automatic FAQ response, medical record input assistance, educational dialogue systems, customer support, etc., suitable for various scenarios triggering automatic detection of useful information.

[0044] The processing unit can infer the user's emotions and determine the priority of processing input based on the inferred emotions. For example, the processing unit infers the user's emotions. User emotions include, for example, anxiety, relaxation, confusion, etc., but are not limited to these examples. For example, when the user is anxious, the processing unit prioritizes processing important input. For example, when the user is relaxed, the processing unit processes input in the usual order. For example, when the user is confused, the processing unit processes input with appropriate priority. Thus, by determining the priority of input based on the user's emotions, important input can be prioritized. Emotion inference can be achieved, for example, through an emotion engine or generative AI. Generative AI can be text-generating AI (such as LLM) or multimodal generative AI, but is not limited to these examples. Some or all of the above processing in the processing unit can be implemented by AI, or AI can be omitted. For example, the processing unit can input the user's emotions into the AI, which then performs emotion inference. Specifically, this processing department extracts multi-dimensional feature vectors from text data, voice data, and input operation logs (such as key press intervals, correction counts, and input interruption times) received from the user input interface. These vectors are then input into an AI model for sentiment inference (such as a BERT-based sentiment classifier or a multimodal Transformer). Examples of inputs to the AI ​​model include "Input text: 'Please process as soon as possible'", "Input speed: 2.0 characters / second", and "Input interruption: 0 times". The AI ​​model outputs sentiment labels (such as anxiety, relaxation, and confusion) and confidence scores (such as anxiety 0.85, relaxation 0.10, and confusion 0.05) from these inputs. Examples of outputs include "Sentence: Anxiety (0.85)" or "Sentence: Confusion (0.60)". Based on the AI ​​model's output, this processing department executes an input processing priority decision algorithm (such as importance scoring and priority queues). For example, when judging "anxiety", important inputs (such as missing proper nouns or interrogative sentences) are prioritized; when judging "relaxation", the usual order is followed; and when judging "confusion", the AI ​​automatically adjusts the priority. The AI ​​model is trained using input datasets with sentiment labels or data with input priority labels, applying cross-entropy loss or reinforcement learning. Subsequent processing reflects the priority decision results in the processing order of the input-assisted workflow or the data transmission order in the parsing department. Unlike traditional simple processing orders or static rules, this processing department achieves dynamic priority control through AI multi-dimensional feature analysis and sentiment inference. The technical effect is that this system can process inputs based on the user's psychological state and urgency, preventing the omission of important inputs, improving conversation efficiency, and optimizing user experience. Applicable fields include general instant messaging applications, business chat, medical case input assistance in medical scenarios, educational dialogue systems, customer support, etc., suitable for various scenarios requiring input priority control.

[0045] The processing department can prioritize the processing of region-specific proper nouns based on the user's geographic location information. For example, considering the user's geographic location, the processing department prioritizes the processing of region-specific proper nouns. Region-specific proper nouns include, but are not limited to, place names and local specialties. For example, when the user is located in a specific region, the processing department prioritizes the processing of proper nouns related to that region. For example, when the user is traveling, the processing department prioritizes the processing of proper nouns related to that travel destination. For example, when the user is in their local area, the processing department prioritizes the processing of local proper nouns. Thus, by prioritizing the processing of region-specific proper nouns based on the user's geographic location information, region-related information can be provided. Some or all of the above processing in the processing department can be implemented using AI, or it can be done without AI. For example, the processing department can input the user's geographic location information into AI, which will then perform proper noun priority processing. Specifically, this processing department obtains the user's terminal's geographic location information in real time (such as GPS coordinates, Wi-Fi location information, region estimation based on IP address, etc.) and extracts it into a location information vector (such as latitude / longitude, region code, movement history, etc.). This processing department combines this location information with the input text (e.g., UTF-8 encoded strings, maximum 512 tokens) and compares it with a dictionary of regionally specific proper nouns (e.g., lists of place names categorized by prefectures, lists of regional specialties, databases of tourist attractions, etc.). Furthermore, the location information vector and input text are fed into an AI model (e.g., a Transformer-based regional relevance inference model, a geographic information embedding model, etc.) to calculate the priority scores of regionally specific proper nouns (e.g., local 0.90, travel destination 0.80, other regions 0.10). Examples of inputs to the AI ​​model include "Location information: Chiyoda Ward, Tokyo" and "Input text: 'Recommended tourist attractions are…'". The AI ​​model outputs a candidate list of proper nouns (e.g., Tokyo Tower 0.85, Senso-ji Temple 0.80, Tokyo Skytree 0.75) and priority labels from these inputs. An example output is "Priority candidate: Tokyo Tower (0.85)". Based on the output of the AI ​​model, this acceptance department prioritizes the acceptance of regionally specific proper nouns and attaches priority information when transmitting data to the parsing and proposal departments. The AI ​​model is trained using input datasets with location information or data labeled with regionally relevant proper nouns, applying cross-entropy loss or ranking learning. Subsequent processing reflects the priority acceptance results of regionally specific proper nouns in the proposal department's candidate generation or option prompts in the selection department. Unlike traditional methods that simply refer to dictionary definitions or use static rules, this acceptance department uses AI to comprehensively analyze location information and context to achieve dynamic priority acceptance of proper nouns. The technical benefits include the system's ability to provide highly accurate regional information matching the user's current location and movement, enhancing user experience through features such as tourist navigation, regionally specific services, and local information support.Applicable fields include general instant messaging applications, travel navigation applications, regional close-to-home services, business chat, educational dialogue systems, etc., and are suitable for various scenarios that require location information linkage input assistance.

[0046] The processing department can analyze users' social media activity and prioritize relevant input. For example, the processing department analyzes users' social media activity, which includes, but is not limited to, post content and number of followers. For example, the processing department prioritizes proper nouns frequently used by users on social media. For example, the processing department prioritizes proper nouns related to topics discussed by users on social media. For example, the processing department prioritizes proper nouns from users' social media activity. Thus, by analyzing users' social media activity, relevant input can be prioritized. Some or all of the above processing in the processing department can be implemented using AI, or it can be done without AI. For example, the processing department can input users' social media activity into AI, which will then perform priority processing of relevant proper nouns. Specifically, this processing department collects posting data (such as text posts, image descriptions, tags, posting time, number of followers, interaction metrics, etc.) from users' associated social media accounts (such as Weibo, image sharing services, video posting services, etc.) in chronological order and extracts it into feature vectors (such as: high-frequency word list, topic category distribution, proper noun frequency, trend score, etc.). This processing department combines these feature values ​​with user input text (e.g., UTF-8 encoded strings, maximum 512 tokens) and inputs them into an AI model (e.g., a Transformer-based topic relevance inference model, graph neural networks, etc.). Examples of inputs to the AI ​​model include "Recent post: 'The newly released movie 〇〇 is very good'", "High-frequency proper nouns: 〇〇, △△", "Input text: 'The lead actress is…'", etc. The AI ​​model outputs a candidate list of relevant proper nouns (e.g., 〇〇 0.90, △△ 0.80, □□ 0.60) and priority scores from these inputs. Examples of outputs include "Priority candidate: 〇〇 (0.90)" or "Priority candidate: △△ (0.80)", etc. Based on the AI ​​model output, this processing department prioritizes processing proper nouns or topics with high topicality on social media and attaches priority information when transmitting data to the parsing and proposal departments. The AI ​​model is trained using social media post and proper noun relevance label data, applying cross-entropy loss function or ranking learning. The subsequent processing prioritizes the acceptance of proposals, reflecting the results in the candidate generation or selection options of the proposal department. Unlike traditional simple keyword frequency references or static rules, this acceptance department utilizes AI-powered social media activity analysis and contextual integration to dynamically prioritize proper nouns. The technical benefits include providing highly accurate information that aligns with users' latest interests, concerns, and trends, thus enhancing engaging conversations, improving information freshness, and optimizing user experience. Applicable areas include general instant messaging applications, SNS-linked chat, marketing support, educational dialogue systems, customer support, and more, suitable for various scenarios requiring social media-linked input assistance.

[0047] The parsing unit can infer the user's emotions and adjust the precision of contextual analysis based on the inferred emotions. For example, the parsing unit infers the user's emotions. User emotions may include, for example, relaxation, urgency, confusion, etc., but are not limited to these examples. For example, the parsing unit performs detailed contextual analysis when the user is relaxed. For example, the parsing unit performs simplified contextual analysis when the user is in urgency. For example, the parsing unit performs high-precision contextual analysis when the user is confused. Thus, by adjusting the precision of contextual analysis based on the user's emotions, more suitable parsing results can be provided. Emotion inference can be achieved, for example, through an emotion engine or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the parsing unit can be implemented using AI, or AI may not be used. For example, the parsing unit can input the user's emotions into the AI, which will then perform the emotion inference. Specifically, this analysis unit extracts multi-dimensional feature vectors from text data (e.g., UTF-8 encoded strings, maximum 512 tokens), speech data (e.g., 16kHz sampled PCM waveform, maximum 30 seconds), and input operation logs (e.g., key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (e.g., software keyboard, speech recognition module). These feature vectors are then input into AI models for sentiment inference (e.g., BERT-based sentiment classifiers, multimodal Transformers, etc.), outputting sentiment labels (e.g., relaxed, urgent, confused) and confidence scores (e.g., relaxed 0.70, urgent 0.20, confused 0.10). Examples of AI model inputs include: "Input text: 'The lead actress in that movie is…'", "Input speed: 1.2 words / second", "Input interruption: 3 times", "Speech tone: high", etc. Examples of AI model outputs include: "Emotion: Relaxed (0.70)" or "Emotion: Urgent (0.20)", etc. Based on the AI ​​model output, this parsing department selects a context parsing precision control algorithm (e.g., detailed parsing mode, simplified parsing mode, precise parsing mode, etc.). For example, during relaxed judgment, a detailed parsing is performed using a multi-stage attention mechanism based on a large language model using Transformer; during urgent judgment, a simplified algorithm that extracts only the main features is used; and during confused judgment, a high-precision parsing that strengthens historical reference and knowledge base integration is employed. The AI ​​model is trained using input datasets with sentiment labels or data with parsing precision labels, applying cross-entropy loss function or multi-task learning. AI input examples include "Input text: 'Please process as soon as possible'", "Input speed: 2.0 words / second", "Input interruption: 0 times", etc., and an output example is "Sentence: Urgent (0.85)". Subsequent processing reflects the parsing precision control results in the proper noun candidate generation or proposal department data transmission method. Unlike traditional uniform context parsing, this parsing department achieves dynamic parsing precision control through AI multi-dimensional feature parsing and sentiment inference.The technical benefits include the ability of this system to automatically optimize parsing accuracy to match the user's psychological state and context, thereby improving the reliability of parsing results, optimizing processing speed, and enhancing user experience. Applicable fields include general instant messaging applications, business chat, medical record input assistance in medical scenarios, educational dialogue systems, customer support, and more; it is suitable for various scenarios where parsing accuracy needs to be controlled based on user emotions.

[0048] The parsing unit can improve parsing accuracy by referencing past dialogue history during context parsing. For example, the parsing unit references past dialogue history during context parsing. Past dialogue history includes, but is not limited to, logs of past user conversations. The parsing unit improves context parsing accuracy by referencing past user dialogue history. For example, the parsing unit improves context parsing accuracy based on proper nouns used by the user in the past. The parsing unit improves context parsing accuracy by detecting specific patterns from the user's past dialogue history. Thus, by referring to past dialogue history, context parsing accuracy can be improved. Some or all of the above processing in the parsing unit can be implemented using AI, or AI can be omitted. For example, the parsing unit can input past dialogue history into AI, which will then perform context parsing accuracy improvement. Specifically, this parsing unit retrieves up to 1000 chronologically stored speech logs (JSON format, each speech with a timestamp, content, topic tags, etc.) from the database for each user, and inputs them along with the input text (e.g., UTF-8 encoded strings, maximum 512 tokens) into an AI model (e.g., a Transformer-based historical reference large-scale language model, an LSTM-based temporal analysis model, etc.). This analysis unit extracts features such as a list of high-frequency proper nouns, topic migration patterns, and user-specific language expressions from historical data, and calculates similarity with the input context (e.g., cosine similarity, vector dot product, etc.). Examples of inputs to the AI ​​model include "Input text: 'The leading actress in that movie is…'", "Past statements: 'The movie I watched yesterday was 〇〇'", and "High-frequency proper nouns: 〇〇, △△", etc. The AI ​​model outputs a historical list with relevance scores from these inputs (e.g., statement ID123 (0.92), statement ID87 (0.75)) and a list of candidate proper nouns (e.g., △△0.85, □□0.10, ××0.05). Examples of outputs include "Relevant history: statement ID123 (0.92)" and "Candidate proper nouns: △△ (0.85)", etc. Based on the AI ​​model's output, this analysis unit dynamically adjusts the weights of the context analysis algorithm or the priority of candidate generation. The AI ​​model is trained using historical data paired with correct proper nouns, and parsing accuracy-labeled data with or without historical reference. Weight optimization is performed using cross-entropy loss or triplet loss. Subsequent processing reflects the historical reference results in the proper noun candidate generation or proposal data transmission. Unlike traditional simple keyword retrieval or static historical reference, this parsing department achieves unconventional and high-precision contextual parsing through AI-integrated historical parsing and dynamic weighting in a high-dimensional vector space. The technical effect is that this system can reflect the user's past speaking tendencies and expression patterns, significantly improving the accuracy of proper noun location and conversational coherence. Applicable fields include general instant messaging applications, business chat, FAQ auto-response, medical record assistance, educational dialogue systems, etc., suitable for various text input assistance scenarios where historical reference is important.

[0049] The parsing unit can apply different parsing algorithms based on the dialogue category during context parsing. For example, the parsing unit can apply different parsing algorithms based on the dialogue category during context parsing. Dialogue categories include, but are not limited to, movies, music, sports, etc. For example, the parsing unit can apply movie-related parsing algorithms when dealing with movies. For example, it can apply music-related parsing algorithms when dealing with music. For example, it can apply sports-related parsing algorithms when dealing with sports. Thus, by applying different parsing algorithms based on the dialogue category, parsing accuracy can be improved. Some or all of the above processing in the parsing unit can be implemented by AI, or AI can be omitted. For example, the parsing unit can input the dialogue category into AI, which will then execute the parsing algorithm application. Specifically, this parsing unit inputs the input text (e.g., UTF-8 encoded string, maximum 512 tokens) into a natural language processing model (e.g., a large language model based on Transformer, BERT, LSTM, etc.), and first uses a topic classifier (e.g., a multi-class classifier, trained on supervised data with category labels) to determine the dialogue category (e.g., movies, music, sports, business, etc.). Input examples for the AI ​​model include "Input text: 'Recommended movies are...'" and "Input text: 'Recently popular songs are...'". The AI ​​model outputs category labels (e.g., Movies 0.95, Music 0.03, Sports 0.02). Output examples include "Category: Movies (0.95)" or "Category: Music (0.90)". Based on the category determination results, this parsing department automatically selects the parsing algorithm optimized for each category (e.g., movie category refers to a movie database, music category refers to lyrics / artist dictionaries, sports category refers to project / athlete name dictionaries, etc.), and performs contextual feature extraction and proper noun candidate generation. The AI ​​model is trained using a conversation dataset with category labels or data with category parsing accuracy labels, applying cross-entropy loss function or multi-task learning. Subsequent processing reflects the category parsing results in the candidate generation of the proposal department or the option prompts of the selection department. Unlike traditional uniform parsing algorithms, this parsing department achieves unconventional and high-precision context parsing through AI category determination and dynamic algorithm switching. The technical benefits include the ability of the system to automatically select the appropriate parsing method based on the dialogue content, significantly improving the accuracy of proper noun location and user satisfaction. Applicable areas include general instant messaging applications, automated FAQ responses, medical record assistance, educational dialogue systems, customer support, and more, making it suitable for various scenarios where category parsing is useful.

[0050] The analysis unit can infer the user's emotions and adjust the display of the analysis results based on the inferred emotions. For example, the analysis unit can infer the user's emotions, such as tension, relaxation, and urgency, but is not limited to these examples. For example, when the user is tense, the analysis unit can provide a concise and easily identifiable display. When the user is relaxed, the analysis unit can provide a display containing detailed information. When the user is in urgency, the analysis unit can provide a display highlighting key points. Thus, by adjusting the display of the analysis results according to the user's emotions, an easy-to-view display can be provided for the user. Emotion inference can be achieved, for example, through an emotion engine or generative AI. Generative AI can be text-generating AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the analysis unit can be implemented using AI, or AI can be omitted. For example, the analysis unit can input the user's emotions into the AI, which will then perform the emotion inference. Specifically, this analysis unit extracts multi-dimensional feature vectors from text data (e.g., UTF-8 encoded strings, maximum 512 tokens), voice data (e.g., 16kHz sampled PCM waveform, maximum 30 seconds), and input operation logs (e.g., key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (e.g., software keyboard, speech recognition module). These vectors are then input into AI models for sentiment estimation (e.g., BERT-based sentiment classifiers, multimodal Transformers, etc.), outputting sentiment labels (e.g., tense, relaxed, urgent) and confidence scores (e.g., tense 0.65, relaxed 0.25, urgent 0.10). Examples of AI model inputs include "Input text: 'Please hurry'", "Input speed: 2.5 characters / second", "Input interruption: 0 times", and "Voice tone: low". Examples of AI model outputs include "Emotion: tense (0.65)" or "Emotion: relaxed (0.25)". Based on the AI ​​model output, this analysis department selects the appropriate display control algorithm (e.g., concise display mode, detailed display mode, key point display mode, etc.). For example, when determining tension, only the main proper noun candidates are highlighted in large font; when determining relaxation, supplementary information and relevant history are displayed in detail for each candidate; and when determining emergency, only key point lists are displayed concisely. The AI ​​model is trained using input datasets with sentiment labels or data with display mode satisfaction labels, applying cross-entropy loss function or multi-task learning. AI input examples include "Input text: 'Tell me now'", "Input speed: 3.0 words / second", "Input interruption: 0 times", etc., and output examples include "Sentiment: Emergency (0.80)". Subsequent processing reflects the display mode control results to the user interface rendering module, achieving automatic optimal display. Unlike traditional uniform display modes or static UIs, this analysis department achieves dynamic display mode control through AI multi-dimensional feature analysis and sentiment inference.The technical benefits include the ability of this system to automatically optimize the display of analytical results to match the user's psychological state and context, thereby improving information visibility, preventing misinterpretation, and enhancing the user experience. Applicable areas include general instant messaging applications, business chat, medical record input assistance in medical scenarios, educational dialogue systems, customer support, and various other scenarios requiring display control based on user emotions.

[0051] The parsing unit can perform parsing based on the user's geographic location information during context parsing. For example, the parsing unit considers the user's geographic location information during context parsing. Geographic location information includes, but is not limited to, GPS data, location information services, etc. For example, when the user is located in a specific region, the parsing unit prioritizes parsing information related to that region. For example, when the user is traveling, the parsing unit prioritizes parsing information about the travel destination. For example, when the user is locally, the parsing unit prioritizes parsing information about the local area. Thus, by parsing based on the user's geographic location information, region-related information can be provided. Some or all of the above processing in the parsing unit can be implemented using AI, or AI can be omitted. For example, the parsing unit can input the user's geographic location information into AI, which will then perform the parsing processing. Specifically, this parsing unit acquires the user's terminal's geographic location information (such as GPS coordinates, Wi-Fi location information, region estimation based on IP address, etc.) in real time and extracts it into a location information vector (such as latitude / longitude, region code, movement history, etc.). This parsing unit combines this location information with the input text (e.g., UTF-8 encoded strings, maximum 512 tokens) and compares it with a dictionary of regionally specific proper nouns (e.g., lists of place names categorized by prefectures, lists of regional specialties, tourist attraction databases, etc.). Furthermore, the location information vector and input text are fed into an AI model (e.g., a Transformer-based regional relevance inference model, a geographic information embedding model, etc.) to calculate the priority scores of regionally specific proper nouns (e.g., local 0.90, travel destination 0.80, other regions 0.10). Examples of inputs to the AI ​​model include "Location information: Kita-ku, Osaka" and "Input text: 'Recommended food is…'". The AI ​​model outputs a candidate list of proper nouns (e.g., takoyaki 0.85, okonomiyaki 0.80, kushikatsu 0.75) and priority labels from these inputs. An example output is "Priority candidate: takoyaki (0.85)". This parsing department prioritizes parsing region-specific proper nouns based on the AI ​​model output, and attaches priority information when transmitting data to the proposal and selection departments. The AI ​​model is trained using input datasets with location information or data labeled with region-related proper nouns, applying cross-entropy loss or ranking learning. Subsequent processing reflects the prioritized parsing results of region-specific proper nouns in the candidate generation of the proposal department or the option suggestions in the selection department. Unlike traditional methods that simply refer to a dictionary or use static rules, this parsing department uses AI to comprehensively analyze location information and context, achieving dynamic priority parsing of proper nouns. The technical benefits include providing highly accurate region-related information that matches the user's current location and movement, enhancing user experience in areas such as travel navigation, region-specific services, and local information support. Applicable fields include general instant messaging applications, travel navigation applications, region-specific services, business chat, and educational dialogue systems, suitable for various scenarios requiring location-based input assistance.

[0052] The parsing department can refer to relevant news or trend information during contextual analysis to improve parsing accuracy. For example, the parsing department can refer to relevant news or trend information during contextual analysis. News or trend information includes, but is not limited to, news feeds and trend databases. For example, the parsing department can refer to the latest news to improve the accuracy of contextual analysis. For example, the parsing department can refer to trend information to improve the accuracy of contextual analysis. For example, the parsing department can improve the accuracy of contextual analysis based on relevant news or trend information. Thus, by referring to relevant news or trend information, the accuracy of contextual analysis can be improved. Some or all of the above processing in the parsing department can be implemented using AI, or AI can be used without it. For example, the parsing department can input news or trend information into AI, which will then perform the accuracy improvement. Specifically, this parsing department obtains structured data such as the latest news titles, text, hashtags, and publication times from external news APIs or trend information databases (such as RSS feeds, SNS trend rankings, news article metadata, etc.). This news / trend data is combined with user input text (e.g., UTF-8 encoded strings, maximum 512 tokens) and fed into a natural language processing model (e.g., a Transformer-based news relevance inference model, a BERT-based topic matching model, etc.). Examples of input to the AI ​​model include "Input text: 'Recently popular movies are…'", "News headline: 'New movie 〇〇 released'", "Trend tag: #movie", etc. The AI ​​model outputs a list of relevant news candidates (e.g., News ID123 (0.92), News ID87 (0.75)) and a list of proper noun candidates with trend scores (e.g., 〇〇 0.88, △△ 0.65). Examples of output include "Related news: News ID123 (0.92)", "Property noun candidate: 〇〇 (0.88)", etc. Based on the AI ​​model's output, this parsing department dynamically adjusts the weights or candidate generation priorities of the context parsing algorithm. The AI ​​model is trained using news / trend data paired with correct proper nouns and relevance tag data, applying cross-entropy loss or ranking learning. Subsequent processing involves reflecting the news / trend reference results in the proper noun candidate generation or proposal data transmission. Unlike traditional simple keyword retrieval or static news reference, this analysis department uses AI to integrate and dynamically weight news / trend analysis in a high-dimensional vector space, achieving unconventional and highly accurate contextual analysis. The technical effect is that this system can reflect users' latest interests, concerns, and social trends, significantly improving the accuracy of proper noun positioning and the freshness of conversations. Applicable fields include general instant messaging applications, SNS-linked chat, marketing assistance, educational dialogue systems, customer support, etc., suitable for various text input assistance scenarios where news / trend reference is important.

[0053] The proposal department can infer the user's emotions and adjust the way the proposal is expressed based on these inferences. For example, the proposal department can infer the user's emotions, such as relaxation, urgency, or confusion, but is not limited to these examples. For instance, when the user is relaxed, the proposal department can provide a detailed proposal. When the user is in urgency, the proposal department can provide a concise proposal. When the user is confused, the proposal department can provide an easily understandable proposal. Thus, by adjusting the way the proposal is expressed based on the user's emotions, a more suitable proposal can be achieved. Emotion inference can be achieved, for example, through an emotion engine or generative AI. Generative AI can be text-generating AI (such as LLM) or multimodal generative AI, but is not limited to these examples. Some or all of the above processing in the proposal department can be achieved through AI, or AI can be omitted. For example, the proposal department can input the user's emotions into the AI, which will then perform the emotion inference. Specifically, this proposal extracts multi-dimensional feature vectors from text data (e.g., UTF-8 encoded strings, maximum 512 tokens), voice data (e.g., 16kHz sampled PCM waveform, maximum 30 seconds), and input operation logs (e.g., key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (e.g., software keyboards, speech recognition modules), and inputs these vectors into an AI model for sentiment inference (e.g., BERT-based sentiment classifiers, multimodal Transformers, etc.). The model outputs sentiment labels (e.g., relaxed, urgent, confused) and confidence scores (e.g., relaxed 0.70, urgent 0.20, confused 0.10). Examples of AI model inputs include: "Input text: 'The lead actress in that movie is…'", "Input speed: 1.2 words / second", "Input interruption: 3 times", "Voice tone: high", etc. Examples of AI model outputs include: "Emotion: Relaxed (0.70)" or "Emotion: Urgent (0.20)", etc. Based on the AI ​​model output, this proposal department selects a proposal expression control algorithm (e.g., detailed proposal mode, concise proposal mode, easy-to-understand mode, etc.). For example, during the relaxation judgment, supplementary information and relevant history are displayed in detail for each proper noun candidate; during the urgency judgment, only the main proper noun candidates are concisely suggested; and during the confusion judgment, explanatory text or example sentences are added to each candidate for easy understanding. The AI ​​model is trained using input datasets with sentiment labels or data with proposal expression satisfaction labels, applying cross-entropy loss function or multi-task learning. AI input examples include "Input text: 'Tell me now'", "Input speed: 3.0 words / second", "Input interruption: 0 times", etc., and output examples include "Sentiment: Urgent (0.80)". Subsequent processing reflects the expression control results in the user interface rendering module, achieving automatic optimal proposal expression. Unlike traditional uniform proposal expression or static UI, this proposal department achieves dynamic proposal expression control through AI multi-dimensional feature analysis and sentiment inference. The technical effect is that this system can automatically optimize the proposal expression to match the user's psychological state and context, improving information visibility, preventing misinterpretation, and enhancing user experience.Applicable fields include general instant messaging applications, business chat, medical case input assistance in medical scenarios, educational dialogue systems, customer support, etc., and it is suitable for a variety of scenarios that require the expression of proposals based on the user's emotions.

[0054] The proposal department can adjust the level of detail in proposals based on the importance of proper nouns. For example, the proposal department adjusts the level of detail based on the importance of proper nouns. The importance of proper nouns may include, but is not limited to, frequency and relevance. For example, the proposal department may make proposals with detailed information for important proper nouns. For example, the proposal department may make concise proposals for general proper nouns. For example, the proposal department may make detailed proposals that fit the context for proper nouns that are important in a specific context. Thus, by adjusting the level of detail in proposals based on the importance of proper nouns, important information can be provided to users. Some or all of the above processing in the proposal department can be implemented using AI, or it may not be necessary to use AI. For example, the proposal department can input the importance of proper nouns into AI, which will then perform the adjustment of the proposal's level of detail. Specifically, this proposal department calculates an importance score (e.g., frequency of occurrence, contextual relevance, user selection history weight, etc.) for each candidate proper noun received by the parsing department. AI models (such as Transformer-based importance estimation models and ranking learning models) take candidate vectors of proper nouns (such as 128-dimensional features), historical data (such as the past 100 selection history entries), and context vectors (such as 512-token context features) as input, and output an importance score for each proper noun (such as 0.95, 0.60, 0.20, etc.). Examples of AI model inputs are "Property Noun: △△, Frequency: High, Relevance: 0.92" and "Property Noun: □□, Frequency: Low, Relevance: 0.45". Examples of AI model outputs are "Importance: △△ (0.95)" and "Importance: □□ (0.45)". Based on the importance score, this proposal applies a detail control algorithm (such as adding detailed information, concise display, and context-dependent explanations) to add detailed information such as explanatory text, related news, and image links to important proper nouns, while only providing concise labels for general proper nouns. The AI ​​model is trained using proper nouns and importance-labeled data, user satisfaction feedback, etc., applying cross-entropy loss function or ranking learning. Subsequent processing reflects the detail control results in the candidate suggestion module of the user interface, facilitating user identification of important information. Unlike traditional uniform candidate suggestions, this proposal uses AI importance estimation and dynamic detail control to provide information tailored to user needs. The technical benefits include preventing information overload or underload, reducing confusion caused by misselection and information overload. Applicable areas include general instant messaging applications, automatic FAQ response, medical record input assistance, educational dialogue systems, customer support, etc., suitable for various scenarios requiring information suggestions based on importance.

[0055] The proposal department can apply different proposal algorithms based on the category of the proper noun during the proposal process. For example, the proposal department applies different proposal algorithms based on the category of the proper noun. Categories of proper nouns include, for example, names of people, names of places, names of products, etc., but are not limited to these examples. For instance, when the proper noun involves a movie, the proposal department applies a movie-related proposal algorithm; when the proper noun involves music, it applies a music-related proposal algorithm; and when the proper noun involves sports, it applies a sports-related proposal algorithm. By applying different proposal algorithms based on the category of the proper noun, the accuracy of the proposal can be improved. Some or all of the above processing in the proposal department can be performed using AI, for example, or without AI. For example, the proposal department can input the category of the proper noun into the AI, and the AI ​​can execute the application of the proposal algorithm. Specifically, for each candidate proper noun received from the parsing department, this proposal department assigns a category label (such as name of person, name of place, name of product, movie, music, sports, etc.) using a category determination module (such as a multi-class classifier, a BERT-based category classification model, etc.). Input examples for the AI ​​model include "property term: △△, context: 'leading actress'" and "property term: □□, context: 'tourist attraction'". The AI ​​model outputs category labels (e.g., movie 0.95, music 0.03, sports 0.02). Output examples are "category: movie (0.95)" or "category: music (0.90)". Based on the category determination results, this proposal department automatically selects the proposal algorithm optimized for each category (e.g., movie category refers to the movie database, music category refers to lyrics / artist dictionary, sports category refers to competition event / player name dictionary, etc.), and performs proper noun candidate generation and supplementary information assignment. The AI ​​model is trained using a proper noun dataset with category labels and data with category proposal precision labels, applying the cross-entropy loss function and multi-task learning. In subsequent processing, the proposal results by category will be reflected in the data transmission of the candidate prompt department or selection department. Unlike the application of traditional uniform proposal algorithms, this proposal department achieves unconventional and high-precision proper noun proposals through AI category determination and dynamic algorithm switching. Its technical advantage lies in its ability to automatically select proposal methods based on dialogue content and category, thereby significantly improving the accuracy of proper noun recognition and user satisfaction. Applicable areas include general instant messaging applications, automated FAQ responses, medical record assistance, educational dialogue systems, customer support, and more, making it suitable for various scenarios where categorized proposals are useful.

[0056] The proposal department can infer the user's emotions and adjust the length of the proposal based on the inferred emotions. For example, the proposal department infers the user's emotions, which may include, but are not limited to, relaxation, anxiety, or confusion. For instance, the proposal department provides a longer proposal when the user is relaxed, a shorter proposal when the user is anxious, and an appropriately long proposal when the user is confused. By adjusting the proposal length based on the user's emotions, a more suitable proposal can be achieved. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the proposal department can be performed using AI, or it can be performed without AI. For example, the proposal department can input the user's emotions into the AI, and the AI ​​can perform emotion inference. Specifically, this proposal department extracts multidimensional feature vectors from text data, voice data, and input operation logs received from the user input interface and inputs them into an AI model for emotion inference (such as a BERT-based emotion classifier, a multimodal Transformer, etc.). Input examples for the AI ​​model include "Input text: 'Please process as soon as possible'", "Input speed: 2.5 words / second", and "Input interruption: 0 times". The AI ​​model outputs sentiment labels (e.g., relaxed, anxious, confused) and confidence scores (e.g., relaxed 0.65, anxious 0.30, confused 0.05). Output examples are "Sentiment: Anxious (0.80)" or "Sentiment: Relaxed (0.70)". Based on the output results of the AI ​​model, this proposal department selects proposal length control algorithms (e.g., long text generation, short text generation, summary generation, etc.). When judging relaxation, it automatically generates long text proposals containing detailed explanations and supplementary information; when judging urgency, it automatically generates short text proposals containing only key points; and when judging confusion, it automatically generates proposals of appropriate length that balance comprehensibility and information content. The AI ​​model is trained using an input dataset with sentiment labels and data with proposal length satisfaction labels, applying cross-entropy loss function and sequence length control learning. In subsequent processing, the proposal length control results will be reflected in the user interface rendering module to achieve the optimal proposal length adapted to the user's psychological state and context. Unlike traditional uniform proposal lengths or static UIs, this proposal department achieves dynamic proposal length control through AI's multi-dimensional feature analysis and sentiment inference. The technical advantage lies in its ability to automatically optimize proposal length to suit the user's psychological state and context, thereby improving information visibility, preventing misunderstandings, and enhancing user experience. Applicable areas include general instant messaging applications, business chat, medical case input assistance in healthcare scenarios, educational dialogue systems, customer support, and more, making it suitable for various scenarios where proposal length needs to be controlled based on user sentiment.

[0057] The proposal department can prioritize proposals based on the submission timing of proper nouns. For example, the proposal department prioritizes proposals based on the submission timing of proper nouns. Submission timing includes, but is not limited to, submission date, submission frequency, etc. The proposal department may prioritize recently used proper nouns; prioritize properly used proper nouns in the past; and prioritize properly used proper nouns related to a specific period. By prioritizing proposals based on the submission timing of proper nouns, important information can be provided to users more efficiently. Some or all of the above processing in the proposal department can be performed using AI, or it can be done without AI. For example, the proposal department can input the submission timing of proper nouns into AI, and let AI determine the proposal priority. Specifically, this proposal department retrieves the proper noun usage history (e.g., a JSON structure containing the speaking ID, proper noun, usage date, usage frequency, etc., up to 1000 entries) stored chronologically by user from the database, and extracts submission timing features (e.g., latest usage date, number of uses in the past 30 days, appearance trend by period, etc.) for each proper noun candidate. AI models (such as LSTM-based temporal pattern recognition models and Transformer-based historical analysis models) take these features as input and output priority scores for each proper noun (e.g., latest 0.95, frequent 0.80, past 0.10). Examples of AI model inputs include "Property noun: △△, latest usage: 2024-06-01, frequency: high" and "Property noun: □□, latest usage: 2023-12-15, frequency: low," etc. Examples of AI model outputs are "Priority: △△ (0.95)" and "Priority: □□ (0.10)," etc. Based on the AI ​​model outputs, this proposal department applies priority control algorithms (such as latest priority, frequent priority, period-weighted, etc.) to prioritize proper nouns highly relevant to the user. The AI ​​model is trained using paired historical data and user selection results, data with submission timing labels, and applies cross-entropy loss and ranking learning. In subsequent processing, the priority control results will be reflected in the data transmission of the candidate suggestion department or selection department. Unlike traditional static candidate prompts or simple frequency references, this proposal utilizes AI-powered temporal historical analysis and dynamic priority control to achieve information prompts that adapt to user usage preferences. The technical advantage lies in the system's ability to prioritize information that users recently need or frequently use, significantly improving conversation efficiency and satisfaction. Applicable areas include general instant messaging applications, automated FAQ responses, medical record input assistance, educational dialogue systems, customer support, and various scenarios suitable for historically linked information prompts.

[0058] The proposal department can adjust the order of proposals based on the relevance of proper nouns. For example, the proposal department adjusts the order of proposals based on the relevance of proper nouns. The relevance of proper nouns includes, for example, co-occurrence relationships, relevance scores, etc., but is not limited to such examples. The proposal department, for example, prioritizes proposing proper nouns most relevant to the context; secondly, it proposes generally relevant proper nouns; and finally, it proposes proper nouns with low relevance. By adjusting the proposal order based on the relevance of proper nouns, highly relevant information can be prioritized for users. Some or all of the above processing in the proposal department can be performed using AI, or it can be performed without AI. For example, the proposal department can input the relevance of proper nouns into AI, and the AI ​​can then perform the adjustment of the proposal order. Specifically, this proposal department integrates contextual features received from the parsing department, a list of candidate proper nouns, past dialogue history, knowledge base information, etc., to calculate the relevance score of each proper noun (such as cosine similarity, co-occurrence frequency, context matching degree, etc.). AI models (such as Transformer-based relevance estimation models and graph neural networks) take context vectors (e.g., 512 tokens), candidate vectors (e.g., 5 items × 128 dimensions), and historical vectors (e.g., 100 items × 256 dimensions) as input, and output relevance scores for each proper noun (e.g., 0.90, 0.60, 0.20, etc.). Examples of AI model inputs include "Property noun: △△, Context matching degree: 0.92, Co-occurrence frequency: High" and "Property noun: □□, Context matching degree: 0.45, Co-occurrence frequency: Low," etc. Examples of AI model outputs include "Relevance: △△ (0.90)" and "Relevance: □□ (0.45)," etc. Based on the AI ​​model outputs, this proposal arranges the proper noun candidates in relevance order, prioritizing the most relevant information for the user. The AI ​​model is trained using data with proper noun and relevance labels, user selection history, and applies cross-entropy loss function and ranking learning. In subsequent processing, the relevance ranking results will be reflected in the data transmission of the candidate suggestion section or selection section. Unlike traditional static candidate suggestions or simple keyword matching, this suggestion section uses AI to estimate relevance in a high-dimensional vector space and dynamically control the order, achieving information prompts that adapt to user intent and context. The technical effect is that this system can improve the efficiency of users obtaining the information they need and reduce confusion caused by misselection and information overload. Applicable fields include general instant messaging applications, automatic FAQ response, medical record input assistance, educational dialogue systems, customer support, etc., suitable for various scenarios that emphasize relevance information prompts.

[0059] The selection unit can infer the user's emotions and adjust the selection method accordingly. For example, the selection unit can infer the user's emotions, such as relaxation, anxiety, or confusion, but is not limited to these examples. For instance, when the user is relaxed, the selection unit provides detailed options; when the user is anxious, it provides concise options; and when the user is confused, it provides easy-to-understand options. By adjusting the selection method based on the user's emotions, more suitable options can be provided. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the selection unit can be performed using AI, or it can be performed without AI. For example, the selection unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference. Specifically, this selection unit extracts multi-dimensional feature vectors from text data (such as UTF-8 encoded strings, up to 512 tokens), voice data (such as 16kHz sampled PCM waveforms, up to 30 seconds), and input operation logs (such as key press intervals, correction counts, input interruption times, etc.) received from the user input interface (such as a software keyboard, a speech recognition module), and inputs these vectors into an AI model for sentiment inference (such as a BERT-based sentiment classifier, a multimodal Transformer, etc.). The AI ​​model outputs sentiment labels (such as relaxed, anxious, confused) and confidence scores (such as relaxed 0.70, anxious 0.20, confused 0.10). Input examples for the AI ​​model include "Input text: 'Which one should I choose?'", "Input speed: 1.0 word / second", "Input interruption: 2 times", "Voice tone: calm", etc. Output examples for the AI ​​model are "Emotion: Relaxed (0.70)" or "Emotion: Anxious (0.20)", etc. Based on the output of the AI ​​model, this selection unit selects an option prompt control algorithm (such as detailed option mode, concise option mode, easy-to-understand priority mode, etc.). For example, when making a relaxed decision, detailed supplementary information and relevant history for each option are displayed; when making an urgent decision, only the main option is briefly suggested; when making a confused decision, explanatory text or example sentences are added to each option to facilitate understanding. The AI ​​model is trained using an input dataset with sentiment labels and data with option suggestion satisfaction labels, applying the cross-entropy loss function and multi-task learning. AI input examples include "Input text: 'Want to choose quickly'", "Input speed: 3.0 words / second", "Input interruption: 0 times", etc., and the output example is "Sentiment: Anxious (0.80)". In subsequent processing, the option suggestion control results are reflected in the user interface rendering module to achieve optimal option suggestions that adapt to the user's psychological state and context. Unlike traditional uniform option suggestions or static UIs, this selection department achieves dynamic option suggestion control through AI's multi-dimensional feature analysis and sentiment inference.Its technical advantage lies in its ability to automatically optimize option prompts to match the user's psychological state and context, thereby improving information visibility, preventing misunderstandings, and enhancing user experience. Applicable areas include general instant messaging applications, business chat, medical record input assistance in healthcare settings, educational dialogue systems, customer support, and more, making it suitable for various scenarios where option prompts are controlled by the user's emotional state.

[0060] The selection unit can provide the optimal selection method by referring to the user's past selection history. This past selection history includes, for example, logs of the user's past choices, but is not limited to such examples. The selection unit can provide the optimal options based on the user's past selection history; prioritize methods frequently chosen by the user in the past; or detect specific patterns from the user's past selection history to provide the optimal selection method. By referring to the user's past selection history, the optimal selection method can be provided. Some or all of the above processing in the selection unit can be performed using AI, or it can be performed without AI. For example, the selection unit can input the user's past selection history into AI, and the AI ​​can perform historical analysis. Specifically, this selection unit retrieves selection history data (such as a JSON structure containing option ID, selection time, selection content, selection reason, and contextual information at the time of selection, up to 1000 records) stored chronologically by user from a database, extracts feature vectors from this historical data (such as option category, selection frequency, selection success rate, selection time period distribution, etc.), and inputs them into an AI model (such as an LSTM-based temporal pattern recognition model, a Transformer-based historical analysis model, etc.). Examples of inputs to the AI ​​model include "30 past selections: 20 names, 5 places, 5 products," "selection success rate: names 0.98, places 0.85," and "selection time period: mostly nighttime." The AI ​​model outputs the optimal selection method (e.g., prioritizing names, prioritizing places, and combining product selection) and recommendation scores (e.g., names 0.92, places 0.08) from these inputs. Output examples include "Recommended selection method: names (0.92)" or "Recommended selection method: places (0.75)." Based on the AI ​​model's output, this selection section automatically selects or prioritizes the recommended selection method on the user interface, facilitating efficient user selection. The AI ​​model is trained using paired data of historical data and user satisfaction and selection success rates, applying cross-entropy loss and reinforcement learning. In subsequent processing, the recommended selection method is reflected in the initial settings of the selection section's option suggestion module or selection assistance triggers. Unlike traditional methods that rely on simple historical references or static selection, this selection system utilizes AI-powered multi-dimensional historical analysis and dynamic selection optimization to provide personalized selection method recommendations. The key advantage is that the system automatically recommends selection methods that align with each user's preferences and efficiency goals, thereby reducing errors, increasing speed, and optimizing user experience. Applicable areas include general instant messaging applications, business chat, medical record input assistance, educational dialogue systems, and customer support, making it suitable for various scenarios that provide optimal selection assistance to each user.

[0061] The selection unit can customize options based on the user's current conversation content during selection. For example, the selection unit parses the user's current conversation content. Conversation content includes, for example, the context of the conversation, related topics, etc., but is not limited to these examples. The selection unit may prioritize providing options related to the user's current conversation content; when the user is discussing a specific topic, it may provide options related to that topic; or it may parse the user's current conversation content and customize and provide the optimal options. By customizing options based on the user's current conversation content, more suitable options can be provided. Some or all of the above processing in the selection unit can be performed using AI, or it can be performed without AI. For example, the selection unit can input the user's conversation content into AI, and the AI ​​can perform option customization. Specifically, this selection unit parses text data (such as UTF-8 encoded strings, up to 512 tokens) or voice data (such as 16kHz sampled PCM waveforms, up to 30 seconds) received from the user input interface in real time and inputs it into a natural language processing module (such as large language models based on Transformer, BERT, LSTM, etc.). This selection unit extracts contextual features such as topic category, previous subject / predicate, and missing proper noun positions from the input context to generate relevant option candidates. Input examples for the AI ​​model include "Input text: 'The recommended movie is…'" and "Previous statement: 'The movie I watched yesterday was 〇〇'". The AI ​​model outputs a list of options with relevance scores from the input context (e.g., movie title A (0.85), movie title B (0.10), movie title C (0.05)). Output examples are "Option: Movie title A (0.85)" or "Option: Movie title B (0.10)". Based on the output of the AI ​​model, this selection function prioritizes the options most relevant to the user's current conversation, optimizing the visibility and operability of the options. The AI ​​model is trained using conversation content and option relevance label data, user selection history, and applies ranking learning (e.g., pairwise loss, listwise loss). In subsequent processing, the customized options are reflected in the option prompt module of the user interface, making it easy for users to select according to the context. Unlike traditional static option prompts or simple keyword matching, this selection function achieves unconventional and highly accurate option customization through AI's contextual understanding and dynamic option generation. Its technical advantage lies in its ability to automatically optimize option suggestions to match the user's conversation content and intent, significantly improving selection accuracy and user satisfaction. Applicable areas include general instant messaging applications, automated FAQ responses, medical record input assistance, educational dialogue systems, customer support, and more, making it suitable for various scenarios with context-sensitive option suggestions.

[0062] The selection unit can infer the user's emotions and determine the priority of selections based on the inferred emotions. For example, the selection unit infers the user's emotions. User emotions include, for example, anxiety, relaxation, confusion, etc., but are not limited to these examples. For instance, when the user is anxious, the selection unit prioritizes important options; when the user is relaxed, it provides options in the usual order; and when the user is confused, the AI ​​provides options with appropriate priority. By determining the priority of selections based on the user's emotions, important options can be prioritized. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the selection unit can be performed using AI, or it can be performed without AI. For example, the selection unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference. Specifically, this selection unit extracts multidimensional feature vectors from text data, voice data, and input operation logs (such as key press intervals, number of corrections, input interruption times, etc.) received from the user input interface, and inputs them into an AI model for emotion inference (such as a BERT-based emotion classifier, multimodal Transformer, etc.). Input examples for the AI ​​model include "Input text: 'Want to select as soon as possible'", "Input speed: 2.0 words / second", and "Input interruptions: 0 times". The AI ​​model outputs sentiment labels (e.g., anxiety, relaxation, confusion) and confidence scores (e.g., anxiety 0.85, relaxation 0.10, confusion 0.05) from these inputs. Output examples are "Sentiment: Anxiety (0.85)" or "Sentiment: Confusion (0.60)". Based on the AI ​​model's output, this selection section executes an option priority determination algorithm (e.g., importance scoring, priority queue, etc.). For example, when determining "anxiety", important options (e.g., missing proper nouns, question sentences, etc.) are prioritized; when determining "relaxation", the usual order is followed; and when determining "confusion", the AI ​​automatically adjusts the priority. The AI ​​model is trained using an input dataset with sentiment labels and data with option priority labels, applying cross-entropy loss and reinforcement learning. In subsequent processing, the priority determination results are reflected in the prompting order of the option suggestion module and the display order of the user interface. Unlike traditional simple prompts or static rules, this selection process utilizes AI's multi-dimensional feature analysis and sentiment inference to achieve dynamic priority control. Its technical advantage lies in its ability to provide option suggestions based on the user's psychological state and urgency, preventing the omission of important options, improving conversation efficiency, and optimizing user experience. Applicable areas include general instant messaging applications, business chat, medical record input assistance in healthcare scenarios, educational dialogue systems, customer support, and more, making it suitable for various scenarios requiring option priority control.

[0063] The selection unit can provide the optimal option based on the user's geographic location information during the selection process. For example, the selection unit considers the user's geographic location information when providing the optimal option. Geographic location information includes, but is not limited to, examples such as GPS data and location information services. For example, when the user is located in a specific region, the selection unit prioritizes options related to that region; when the user is traveling, it prioritizes options related to the travel destination; and when the user is locally, it prioritizes local options. By providing the optimal option based on the user's geographic location information, region-related information can be provided. Some or all of the above processing in the selection unit can be performed using AI, or it can be performed without AI. For example, the selection unit can input the user's geographic location information into AI, and the AI ​​can then provide the options. Specifically, this selection unit acquires geographic location information (such as GPS coordinates, Wi-Fi location information, and region estimation based on IP address) from the user's terminal in real time and extracts it into a location information vector (such as latitude / longitude, region code, and movement history). This selection unit combines this location information with the input text (such as a UTF-8 encoded string with up to 512 tokens) and compares it with a dictionary of region-specific options (such as a list of place names categorized by prefecture, a list of regional specialties, a database of tourist attractions, etc.). Furthermore, the location information vector and the input text are fed into an AI model (such as a Transformer-based regional relevance inference model, a geographic information embedding model, etc.) to calculate the priority scores of region-specific options (e.g., local 0.90, travel destination 0.80, other regions 0.10). Examples of inputs to the AI ​​model include "Location information: Kita-ku, Osaka" and "Input text: 'Recommended food is…'". The AI ​​model outputs a candidate list of options (e.g., takoyaki 0.85, okonomiyaki 0.80, kushikatsu 0.75) and priority labels from these inputs. An example output is "Priority candidate: Takoyaki (0.85)". Based on the output of the AI ​​model, this selection unit prioritizes region-specific options, making it easy for users to choose based on their current location or movement. The AI ​​model is trained using an input dataset with location information and region-related option labels, applying cross-entropy loss and ranking learning. In subsequent processing, the priority suggestions for region-specific options are reflected in the user interface's option suggestion module. Unlike traditional methods that rely on simple dictionary references or static rules, this selection process integrates AI's location information with contextual analysis to achieve dynamic option priority suggestions. The technical advantage lies in its ability to provide highly accurate region-related information adapted to the user's current location or movement, enhancing user experience in areas such as travel navigation, region-specific services, and local information assistance. Applicable fields include general instant messaging applications, travel navigation applications, region-specific services, business chat, and educational dialogue systems, making it suitable for various scenarios requiring location-linked selection assistance.

[0064] The selection department can analyze a user's social media activity and propose options during the selection process. For example, the selection department analyzes a user's social media activity. Social media activity includes, but is not limited to, examples such as content posted and number of followers. The selection department may prioritize options frequently used by the user on social media; provide options related to topics discussed by the user on social media; or prioritize relevant options from the user's social media activity. Relevant options can be provided by analyzing the user's social media activity. Some or all of the above processing in the selection department may be performed using AI, or it may not. For example, the selection department can input the user's social media activity into AI, and the AI ​​can then propose options. Specifically, this selection department collects posting data (such as text posts, image descriptions, tags, posting time, number of followers, interaction metrics, etc.) from the user's associated social media accounts (such as Weibo, photo sharing services, video posting services, etc.) in chronological order and extracts it into feature vectors (such as a list of high-frequency words, topic category distribution, option frequency, trend score, etc.). This selection unit combines these features with user input text (such as a UTF-8 encoded string with a maximum of 512 tokens) and inputs it into an AI model (such as a Transformer-based topic relevance inference model, graph neural network, etc.). Examples of AI model inputs include "Recently Posted: 'New movie 〇〇 is very interesting'", "High-Frequency Options: 〇〇, △△", "Input Text: 'The lead actress is…'", etc. The AI ​​model outputs a list of relevant option candidates (such as 〇〇 0.90, △△ 0.80, □□ 0.60) and priority scores from these inputs. Examples of outputs are "Priority Candidate: 〇〇 (0.90)" or "Priority Candidate: △△ (0.80)", etc. Based on the AI ​​model's output, this selection unit prioritizes options or topics with high relevance on social media, making it easier for users to choose based on their latest interests, concerns, or trends. The AI ​​model is trained using social media posting and option relevance label data, applying cross-entropy loss and ranking learning. In subsequent processing, the priority suggestion results will be reflected in the display order of the option suggestion module or user interface. Unlike traditional simple keyword frequency references or static rules, this selection tool utilizes AI-powered social media activity analysis and contextual integration to achieve dynamic option prioritization suggestions. Its technical advantage lies in its ability to provide highly accurate information tailored to the user's latest interests, concerns, or trends, thereby enhancing engaging conversations, keeping information fresh, and optimizing the user experience. Applicable areas include general instant messaging applications, SNS-integrated chat, marketing support, educational dialogue systems, and customer support, making it suitable for various scenarios involving social media-integrated selection assistance.

[0065] The reference unit can infer the user's emotions and select reference data based on the inferred emotions. For example, the reference unit infers the user's emotions. User emotions include, for example, relaxation, anxiety, confusion, etc., but are not limited to these examples. For example, when the user is relaxed, the reference unit refers to detailed data; when the user is anxious, it refers to simplified data; and when the user is confused, it refers to high-precision data. By selecting reference data based on the user's emotions, more suitable data can be provided. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the reference unit can be performed using AI, or it can be performed without AI. For example, the reference unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference. Specifically, this reference unit extracts multi-dimensional feature vectors from text data (such as UTF-8 encoded strings, up to 512 tokens), speech data (such as 16kHz sampled PCM waveforms, up to 30 seconds), and input operation logs (such as key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (such as software keyboards, speech recognition modules), and inputs these vectors into an AI model for sentiment inference (such as a BERT-based sentiment classifier, a multimodal Transformer, etc.). The AI ​​model outputs sentiment labels (such as relaxed, anxious, confused) and confidence scores (such as relaxed 0.70, anxious 0.20, confused 0.10). Input examples for the AI ​​model include "Input text: 'Want to know more details'", "Input speed: 1.0 word / second", "Input interruption: 2 times", "Voice tone: calm", etc. Output examples for the AI ​​model are "Emotion: Relaxed (0.70)" or "Emotion: Anxious (0.20)", etc. Based on the output of the AI ​​model, this reference unit selects data selection algorithms (such as detailed data priority, simple data priority, high-precision data priority, etc.). When making relaxed judgments, it refers to detailed relevant data (such as multiple references, detailed statistical data, supplementary images, etc.); when making urgent judgments, it refers to simple data containing only key points (such as abstracts, main values, bullet points, etc.); and when making confused judgments, it prioritizes high-confidence data (such as official databases, verified information, etc.). The AI ​​model is trained using an input dataset with sentiment labels and data with data selection satisfaction labels, applying the cross-entropy loss function and multi-task learning. AI input examples include "Input text: 'Want to know the result as soon as possible'", "Input speed: 3.0 words / second", "Input interruption: 0 times", etc., with an output example of "Sentiment: Anxious (0.80)". In subsequent processing, the data selection results are reflected in the data acquisition module of the reference unit or the data transmission method of the parsing unit. Unlike traditional uniform data references or static rules, this reference unit achieves dynamic data selection control through AI's multi-dimensional feature analysis and sentiment inference.Its technical advantage lies in its ability to automatically optimize data references to align with the user's psychological state and context, thereby improving information reliability, processing speed, and user experience. Applicable areas include general instant messaging applications, business chat, medical case reference assistance, educational dialogue systems, customer support, and more, making it suitable for various scenarios where data references are controlled based on user emotions.

[0066] The reference unit can optimize the parsing algorithm by referring to past dialogue history during reference. For example, the reference unit references past dialogue history. Past dialogue history includes, for example, user's past conversation logs, but is not limited to such examples. The reference unit optimizes the parsing algorithm by referring to the user's past dialogue history; optimizes the parsing algorithm based on proper nouns used by the user in the past; and optimizes the parsing algorithm by detecting specific patterns from the user's past dialogue history. By referring to past dialogue history, the accuracy of the parsing algorithm can be improved. Some or all of the above processing in the reference unit can be performed using AI, or it can be performed without AI. For example, the reference unit can input past dialogue history into AI, and the AI ​​can perform the optimization of the parsing algorithm. Specifically, this reference unit retrieves up to 1000 speech logs (in JSON format, each speech with a timestamp, speech content, topic tags, etc.) stored chronologically by user from the database, and inputs them along with input text (such as a UTF-8 encoded string with up to 512 tokens) into an AI model (such as a Transformer-based historical reference large-scale language model, an LSTM-based temporal parsing model, etc.). This reference unit extracts features such as a list of high-frequency proper nouns, topic migration patterns, and user-specific language expressions from historical data, and calculates similarity with the input context (e.g., cosine similarity, vector dot product). Input examples for the AI ​​model include "Input text: 'The recommended movie is…'", "Past statements: 'The movie I watched yesterday was 〇〇'", and "High-frequency proper nouns: 〇〇, △△". The AI ​​model outputs a historical list with relevance scores from these inputs (e.g., statement ID123 (0.92), statement ID87 (0.75)) and a list of candidate proper nouns (e.g., △△0.85, □□0.10, ××0.05). Output examples include "Relevant history: statement ID123 (0.92)" and "Candidate proper nouns: △△ (0.85)". Based on the output of the AI ​​model, this reference unit dynamically adjusts the weights of the parsing algorithm and the priority of candidate generation. The AI ​​model is trained using historical data paired with correct proper nouns and parsing accuracy labels indicating whether or not historical references are used. Weight optimization is achieved through cross-entropy loss and triplet loss. In subsequent processing, the historical reference results are reflected in the parsing department's algorithm selection or candidate generation module. Unlike traditional simple keyword retrieval or static historical reference, this reference department achieves unconventional and high-precision parsing algorithm optimization through AI's historical integration analysis and dynamic weight adjustment in a high-dimensional vector space. Its technical advantage lies in its ability to reflect users' past speaking tendencies and expression patterns during parsing, significantly improving proper noun recognition accuracy and conversational coherence. Applicable fields include general instant messaging applications, business chat, FAQ auto-response, medical record assistance, and educational dialogue systems, making it suitable for various text input assistance scenarios where historical reference is crucial.

[0067] The reference unit can infer the user's emotions and adjust the frequency of reference based on the inferred emotions. For example, the reference unit infers the user's emotions. User emotions include, for example, tension, relaxation, and anxiety, but are not limited to these examples. The reference unit, for instance, references data frequently when the user is tense; references data at a normal frequency when the user is relaxed; and references data at the minimum frequency when the user is anxious. By adjusting the frequency of reference based on the user's emotions, more suitable data can be provided. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the reference unit can be performed using AI, or it can be performed without AI. For example, the reference unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference. Specifically, this reference unit extracts multi-dimensional feature vectors from text data (such as UTF-8 encoded strings, up to 512 tokens), speech data (such as 16kHz sampled PCM waveforms, up to 30 seconds), and input operation logs (such as key press intervals, correction counts, input interruption times, etc.) received from user input interfaces (such as software keyboards, speech recognition modules), and inputs these vectors into an AI model for sentiment inference (such as a BERT-based sentiment classifier, a multimodal Transformer, etc.). The AI ​​model outputs sentiment labels (such as tense, relaxed, anxious) and confidence scores (such as tense 0.65, relaxed 0.25, anxious 0.10). Input examples for the AI ​​model include "Input text: 'What to do…'", "Input speed: 0.8 characters / second", "Input interruption: 5 times", "Voice tone: high", etc. Output examples for the AI ​​model are "Emotion: tense (0.65)" or "Emotion: anxious (0.30)", etc. Based on the output of the AI ​​model, this reference unit selects a reference frequency control algorithm (such as high-frequency reference mode, normal reference mode, low-frequency reference mode, etc.). During tense judgments, it references multiple data sources in real-time at high frequency; during relaxed judgments, it references data at normal times; and during urgent judgments, it references only the minimum necessary data. The AI ​​model is trained using an input dataset with sentiment labels and data labeled with reference frequency satisfaction, applying the cross-entropy loss function and multi-task learning. AI input examples include "Input text: 'Want to know immediately'", "Input speed: 3.0 words / second", "Input interruption: 0 times", etc., with an output example of "Sentence: Anxious (0.80)". In subsequent processing, the reference frequency control results are reflected in the data acquisition module or the data transmission method of the parsing unit. Unlike traditional uniform data reference frequencies or static rules, this reference unit achieves dynamic reference frequency control through multi-dimensional feature analysis and sentiment inference of AI. Its technical effect is that this system can automatically optimize the data reference frequency to adapt to the user's psychological state and context, thereby improving information freshness, optimizing communication load, and enhancing user experience.Applicable fields include general instant messaging applications, business chat, medical case reference assistance, educational dialogue systems, customer support, etc., and it is suitable for various scenarios where the reference frequency is controlled according to the user's emotions.

[0068] The reference unit can weight reference data based on the submission timing of the dialogue during reference. For example, the reference unit weights the reference data based on the submission timing of the dialogue. The submission timing of the dialogue includes, for example, submission date, submission frequency, etc., but is not limited to such examples. The reference unit may, for example, prioritize referencing the most recent dialogue history; appropriately weight referencing past dialogue history; and weight referencing dialogue history related to a specific period. By weighting the reference data based on the submission timing of the dialogue, more suitable data can be provided. Some or all of the above processing in the reference unit can be performed using AI, or it can be performed without AI. For example, the reference unit can input the submission timing of the dialogue into AI, and let AI perform data weighting. Specifically, this reference unit retrieves dialogue history data (such as a JSON structure containing message ID, message content, submission date, submission frequency, etc., up to 1000 records) stored chronologically by user from the database, and extracts submission timing features (such as latest usage date, usage frequency in the past 30 days, and occurrence trends by period) for each dialogue history record. AI models (such as LSTM-based temporal pattern recognition models and Transformer-based historical analysis models) use these features as input and output weighted scores for each historical data point (e.g., latest 0.95, frequent 0.80, past 0.10). Examples of AI model inputs include "Message ID: 123, latest usage: 2024-06-01, frequency: high" and "Message ID: 456, latest usage: 2023-12-15, frequency: low," etc. Examples of AI model outputs are "Weighted: Message ID 123 (0.95)" and "Weighted: Message ID 456 (0.10)," etc. Based on the AI ​​model outputs, this reference unit applies weighted control algorithms (such as latest priority, frequent priority, and period-based weighting) to prioritize historical data highly relevant to the user. The AI ​​model is trained using paired data of historical data and user selection results, data with submission timing labels, and applies cross-entropy loss and ranking learning. In subsequent processing, the weighted control results are reflected in the data acquisition module or the data transmission of the parsing unit. Unlike traditional static historical references or simple frequency references, this reference system utilizes AI-powered temporal historical analysis and dynamic weighted control to achieve information references that adapt to user usage preferences. Its technical advantage lies in prioritizing information that users recently need or frequently use, significantly improving conversation efficiency and satisfaction. Applicable areas include general instant messaging applications, automated FAQ responses, medical record reference assistance, educational dialogue systems, customer support, and more, making it suitable for various scenarios involving historically linked information references.

[0069] The candidate suggestion unit can infer the user's emotions and adjust the suggestion method based on the inferred emotions. For example, the candidate suggestion unit infers the user's emotions. User emotions include, for example, relaxation, anxiety, confusion, etc., but are not limited to these examples. For example, when the user is relaxed, the candidate suggestion unit suggests detailed candidates; when the user is anxious, it suggests concise candidates; when the user is confused, it suggests easy-to-understand candidates. By adjusting the suggestion method based on the user's emotions, more suitable candidates can be provided. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the candidate suggestion unit can be performed using AI, or it can be performed without AI. For example, the candidate suggestion unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference.

[0070] The candidate suggestion unit can suggest the optimal candidate option by referring to the user's past selection history. This reference includes, for example, a log of the user's past choices, but is not limited to such examples. The unit may suggest the optimal candidate option based on the user's past selection history; prioritize frequently selected options; or detect specific patterns in the user's past selection history to suggest the optimal candidate option. By referring to the user's past selection history, optimal candidate options can be provided. Some or all of the above processing in the candidate suggestion unit can be performed using AI, or it can be performed without AI. For example, the candidate suggestion unit can input the user's past selection history into AI, and the AI ​​can perform historical analysis.

[0071] The candidate suggestion unit can infer the user's emotions and determine the priority of candidates based on the inferred emotions. For example, the candidate suggestion unit infers the user's emotions. User emotions include, for example, anxiety, relaxation, confusion, etc., but are not limited to these examples. For example, when the user is anxious, the candidate suggestion unit prioritizes important candidates; when the user is relaxed, it suggests candidates in the usual order; when the user is confused, the AI ​​suggests candidates with appropriate priority. By determining the priority of candidates based on the user's emotions, important candidates can be provided first. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the candidate suggestion unit can be performed using AI, or it can be performed without AI. For example, the candidate suggestion unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference.

[0072] The candidate suggestion unit can suggest the optimal candidate option based on the user's geographical location information when providing suggestions. For example, the candidate suggestion unit considers the user's geographical location information when suggesting the optimal candidate option. Geographical location information includes, but is not limited to, examples such as GPS data and location information services. For example, when the user is located in a specific region, the candidate suggestion unit prioritizes suggestions related to that region; when the user is traveling, it prioritizes suggestions related to the travel destination; and when the user is locally, it prioritizes suggestions related to the local area. By suggesting the optimal candidate option based on the user's geographical location information, region-related information can be provided. Some or all of the above processing in the candidate suggestion unit can be performed using AI, or it can be performed without AI. For example, the candidate suggestion unit can input the user's geographical location information into AI, and the AI ​​can then perform the suggestion of candidate options.

[0073] The response unit can infer the user's emotions and adjust the response method accordingly. For example, the response unit infers the user's emotions. User emotions include, for example, relaxation, anxiety, confusion, etc., but are not limited to these examples. The response unit, for example, provides a detailed response method when the user is relaxed; a concise response method when the user is anxious; and an easy-to-understand response method when the user is confused. By adjusting the response method according to the user's emotions, a more appropriate response can be achieved. Emotion inference can be achieved through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (such as LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the response unit can be performed using AI, or it can be performed without AI. For example, the response unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference.

[0074] The feedback unit can provide the optimal feedback method by referring to the user's past selection history. For example, the feedback unit refers to the user's past selection history. Past selection history includes, for example, a log of the user's past choices, but is not limited to such examples. The feedback unit provides the optimal feedback method based on the user's past selection history; prioritizes methods frequently chosen by the user in the past; or detects specific patterns from the user's past selection history to provide the optimal feedback method. By referring to the user's past selection history, the optimal feedback method can be provided. Some or all of the above processing in the feedback unit can be performed using AI, or it can be performed without AI. For example, the feedback unit can input the user's past selection history into AI, and the AI ​​can perform historical analysis.

[0075] The response unit can infer the user's emotions and determine the priority of responses based on the inferred emotions. For example, the response unit can infer the user's emotions. User emotions include, for example, anxiety, relaxation, confusion, etc., but are not limited to these examples. For example, when the user is anxious, the response unit prioritizes important responses. For example, when the user is relaxed, the response unit responds in the usual order. For example, when the user is confused, the response unit uses AI to respond with appropriate priority. Thus, by determining the priority of responses based on the user's emotions, important responses can be prioritized. Emotion inference can be achieved, for example, through emotion inference functions such as emotion engines or generative AI. Generative AI can be text generation AI (e.g., LLM) or multimodal generation AI, but is not limited to these examples. Some or all of the above processing in the response unit can be implemented using AI, or it can be implemented without AI. For example, the response unit can input the user's emotions into the AI, and the AI ​​can perform emotion inference.

[0076] The response department can provide the optimal response method based on the user's geographic location information during the response process. For example, the response department can consider the user's geographic location information to provide the optimal response method. Geographic location information includes, but is not limited to, GPS data, location information services, etc. For example, when the user is located in a specific region, the response department provides a response method related to that region. For example, when the user is traveling, the response department provides a response method related to the travel destination. For example, when the user is locally, the response department provides a local response method. Thus, by providing the optimal response method based on the user's geographic location information, region-related information can be provided. Some or all of the above processing in the response department can be implemented using AI, or it can be implemented without AI. For example, the response department can input the user's geographic location information into AI, and the AI ​​can then provide the response method.

[0077] The system involved in this embodiment is not limited to the examples described above. For example, various modifications can be made as described below.

[0078] When receiving user input, the processing unit can analyze the user's input speed and adjust the timing of input assistance. For example, when the user inputs at a slower speed than usual, the processing unit can immediately activate input assistance. Conversely, when the user inputs at a high speed, the processing unit can delay the activation of input assistance. Furthermore, when the user inputs at a constant speed, the processing unit can activate input assistance at the usual time. Thus, input assistance can be provided at appropriate times based on the user's input speed.

[0079] The parsing unit can retrieve relevant images or videos based on user input to improve the accuracy of contextual analysis. For example, when a user inputs "The lead actress in the movie is...", the parsing unit can retrieve movie posters or trailer videos for contextual analysis. Similarly, when a user inputs "The famous sights of the travel destination are...", the parsing unit can retrieve photos or sightseeing videos of the destination for contextual analysis. Furthermore, when a user inputs "The name of the new gadget is...", the parsing unit can retrieve images or review videos of that gadget for contextual analysis. Thus, by utilizing relevant images or videos, the accuracy of contextual analysis can be improved.

[0080] The proposal department can retrieve relevant news reports or blog posts based on user input and use them for proper noun proposals. For example, when a user enters "The leading actress in the recent movie is...", the proposal department can retrieve news reports or blog posts related to the recent movie and use them for proper noun proposals. Similarly, when a user enters "The name of the new smartphone is...", the proposal department can retrieve news reports or blog posts related to the new smartphone and use them for proper noun proposals. Furthermore, when a user enters "The name of a famous tourist destination is...", the proposal department can retrieve news reports or blog posts related to the tourist destination and use them for proper noun proposals. Therefore, by utilizing relevant news reports or blog posts, the accuracy of proper noun proposals can be improved.

[0081] The selection menu can customize the display order of options based on the user's selection history. For example, it can prioritize displaying proper nouns that the user has frequently selected in the past. Additionally, the display order can be adjusted based on the relevance of the proper nouns selected in the past. Furthermore, the display order can be customized based on the category of the proper nouns selected in the past. Therefore, by customizing the display order of options based on the user's selection history, more suitable options can be provided.

[0082] The service department can infer a user's emotions and customize input assistance content based on these inferences. For example, when a user is anxious, the service department can provide concise and quick input assistance. Conversely, when a user is relaxed, the service department can provide detailed and thoughtful input assistance. Furthermore, when a user is confused, the service department can provide easy-to-understand and considerate input assistance. Thus, by customizing input assistance content according to the user's emotions, more appropriate assistance can be provided.

[0083] The parsing unit can retrieve relevant audio data based on user input to improve the accuracy of contextual analysis. For example, when a user inputs "A part of a famous speech is...", the parsing unit can retrieve the audio data of that speech and use it for contextual analysis. Similarly, when a user inputs "The lyrics of a popular song are...", the parsing unit can retrieve the audio data of that song and use it for contextual analysis. Furthermore, when a user inputs "A specific episode of a particular podcast is...", the parsing unit can retrieve the audio data of that podcast and use it for contextual analysis. Thus, by utilizing relevant audio data, the accuracy of contextual analysis can be improved.

[0084] The proposal department can retrieve relevant books or papers based on user input and use them for proper noun proposals. For example, when a user enters "The name of a famous author is...", the proposal department can retrieve books or papers related to that author and use them for proper noun proposals. Similarly, when a user enters "The name of a specific scientist is...", the proposal department can retrieve books or papers related to that scientist and use them for proper noun proposals. Furthermore, when a user enters "The name of a historical figure is...", the proposal department can retrieve books or papers related to that figure and use them for proper noun proposals. Therefore, by utilizing relevant books or papers, the accuracy of proper noun proposals can be improved.

[0085] The selection panel can infer the user's emotions and adjust the display of options based on these inferences. For example, when a user is anxious, the selection panel can provide simple and easily identifiable options. Conversely, when a user is relaxed, the selection panel can provide options with detailed information. Furthermore, when a user is confused, the selection panel can provide easy-to-understand and considerate options. Thus, by adjusting the display of options according to the user's emotions, more appropriate options can be provided.

[0086] The parsing unit can retrieve relevant statistical data or charts based on user input to improve the accuracy of contextual analysis. For example, when a user inputs "The data for economic indicators is...", the parsing unit can retrieve statistical data or charts related to that economic indicator and use them for contextual analysis. Similarly, when a user inputs "The chart for market trends is...", the parsing unit can retrieve statistical data or charts related to that market trend and use them for contextual analysis. Furthermore, when a user inputs "The data for demographics is...", the parsing unit can retrieve statistical data or charts related to that demographic statistic and use them for contextual analysis. Thus, by utilizing relevant statistical data or charts, the accuracy of contextual analysis can be improved.

[0087] The proposal department can infer users' emotions and customize proposal content based on these inferences. For example, when a user is anxious, the department can provide a concise and quick proposal. Conversely, when a user is relaxed, the department can provide a detailed and thoughtful proposal. Furthermore, when a user is confused, the department can provide an easy-to-understand and considerate proposal. Thus, by customizing proposal content according to users' emotions, more suitable proposals can be provided.

[0088] The following is a brief description of the processing flow of the implementation method.

[0089] Step 1: The processing department receives user input. User input may include, but is not limited to, text input, voice input, etc. For example, if a user cannot recall a proper noun while conversing using an instant messaging application, the processing department will invoke AI input assistance.

[0090] Step 2: The parsing department analyzes the context based on the input received by the receiving department. Context analysis may include, but is not limited to, the context of the conversation, related topics, etc. The parsing department may refer to past dialogue history or a general database to analyze the context.

[0091] Step 3: The proposal department proposes proper nouns based on the context analyzed by the analysis department. Proposals for proper nouns may include, but are not limited to, names of people, places, and products. The proposal department may provide multiple options.

[0092] Step 4: The selection section allows the user to choose the proper nouns proposed in the proposal section. The selection section reflects the user's choice. Therefore, even if the user cannot recall a proper noun during the conversation, they can continue the dialogue smoothly.

[0093] The specific processing unit 290 sends the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires voice representing the user's input to the result of the specific processing. The control unit 46A sends the voice data representing the user's input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0094] Data generation model 58 is what is known as generative AI (Artificial Intelligence). An example of data generation model 58 includes ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> Generative AI, such as data generation model 58, is obtained by deep learning through a neural network. The data generation model 58 is input with a prompt containing instructions, and with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the above-described specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts without instructions; in this case, the data generation model 58 can output inference results from prompts without instructions. The data processing device 12, etc., includes multiple data generation models 58, including AI other than generative AI. AI other than generative AI includes, but is not limited to, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes. Furthermore, AI can also act as an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to this example. Moreover, processing performed by AI, including generative AI, can be replaced by rule-based processing, and vice versa.

[0095] Furthermore, the processing performed by the aforementioned data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart device 14 or external devices, and the smart device 14 acquires or collects information required for processing from the data processing device 12 or external devices.

[0096] Each of the aforementioned elements, including the receiving unit, parsing unit, proposal unit, and selection unit, can be implemented, for example, in at least one of the smart device 14 and the data processing device 12. For instance, the receiving unit, implemented by the control unit 46A of the smart device 14, is used to receive user input. The parsing unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for parsing the context. The proposal unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for proposing proper nouns. The selection unit, for example, is implemented by the control unit 46A of the smart device 14, for reflecting the user's selection. The correspondence between the various units and the device or control unit is not limited to the above examples and can be varied.

[0097] [Second Implementation] Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0098] like Figure 3 As shown, the data processing system 210 includes a data processing device 12 and smart glasses 214. One example of the data processing device 12 is a server.

[0099] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0100] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0101] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0102] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0103] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0104] Figure 4 An example of the main functions of the data processing device 12 and the smart glasses 214 is shown. Figure 4 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0105] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0106] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0107] In the smart glasses 214, specific processing is performed by the processor 46. A specific processing program 60 is stored in the memory 50. The processor 46 reads the specific processing program 60 from the memory 50 and executes the read specific processing program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific processing program 60 executed on the RAM 48. Furthermore, the smart glasses 214 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0108] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0109] The specific processing unit 290 sends the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires voice input representing the user's input to the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0110] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0111] The data processing system 210 of the second embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 210 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or external devices, and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or external devices.

[0112] Each of the aforementioned elements, including the receiving unit, parsing unit, proposal unit, and selection unit, can be implemented, for example, in at least one of the smart glasses 214 and the data processing device 12. For instance, the receiving unit, implemented by the control unit 46A of the smart glasses 214, is used to receive user input. The parsing unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for parsing the context. The proposal unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for proposing proper nouns. The selection unit, for example, is implemented by the control unit 46A of the smart glasses 214, for reflecting the user's selection. The correspondence between the various units and the device or control unit is not limited to the above examples and can be varied.

[0113] [Third Implementation] Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0114] like Figure 5 As shown, the data processing system 310 includes a data processing device 12 and a head-mounted terminal 314. One example of the data processing device 12 is a server.

[0115] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0116] The head-mounted terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0117] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0118] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0119] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0120] Figure 6 An example of the main functions of the data processing device 12 and the head-mounted terminal 314 is shown. Figure 6 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0121] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0122] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0123] In the head-mounted terminal 314, specific processing is performed by the processor 46. A specific program 60 is stored in the memory 50. The processor 46 reads the specific program 60 from the memory 50 and executes the read specific program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific program 60 executed on the RAM 48. Furthermore, the head-mounted terminal 314 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0124] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0125] The specific processing unit 290 sends the result of the specific processing to the head-mounted terminal 314. In the head-mounted terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input to the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0126] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0127] The data processing system 310 of the third embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 310 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the head-mounted terminal 314, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the head-mounted terminal 314. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the head-mounted terminal 314 or external devices, and the head-mounted terminal 314 acquires or collects information required for processing from the data processing device 12 or external devices.

[0128] Each of the aforementioned elements, including the receiving unit, parsing unit, proposal unit, and selection unit, can be implemented, for example, in at least one of the head-mounted terminal 314 and the data processing device 12. For example, the receiving unit, implemented by the control unit 46A of the head-mounted terminal 314, is used to receive user input. The parsing unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for parsing the context. The proposal unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for proposing proper nouns. The selection unit, for example, is implemented by the control unit 46A of the head-mounted terminal 314, for reflecting the user's selection. The correspondence between the various units and the device or control unit is not limited to the above examples and can be varied.

[0129] [Fourth Implementation] Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0130] like Figure 7 As shown, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0131] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0132] Robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control object 443. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and control object 443 are also connected to the bus 52.

[0133] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0134] The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS image sensor or a CCD image sensor, used to photograph the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0135] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0136] The controlled object 443 includes a display device, LEDs for the eyes, and motors for driving the arms, hands, and feet. The posture and movements of the robot 414 are controlled by controlling the motors for the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. In addition, facial expressions of the robot 414 can also be expressed by controlling the illumination state of the LEDs for the robot 414's eyes.

[0137] Figure 8 An example of the main functions of the data processing device 12 and the robot 414 is shown. For example... Figure 8 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0138] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0139] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0140] In robot 414, specific processing is performed by processor 46. A specific program 60 is stored in memory 50. Processor 46 reads the specific program 60 from memory 50 and executes the read specific program 60 on RAM 48. Specific processing is achieved by processor 46 acting as control unit 46A based on the specific program 60 executed on RAM 48. Furthermore, robot 414 may also have the same data generation model and emotion-specific model as data generation model 58 and emotion-specific model 59, and use these models to perform the same processing as specific processing unit 290.

[0141] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0142] The specific processing unit 290 sends the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input regarding the result of the specific processing. The control unit 46A sends the voice data representing the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0143] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0144] The data processing system 410 of the fourth embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 410 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or external devices, and the robot 414 acquires or collects information required for processing from the data processing device 12 or external devices.

[0145] Each of the aforementioned elements, including the receiving unit, parsing unit, proposal unit, and selection unit, can be implemented, for example, in at least one of the robot 414 and the data processing device 12. For instance, the receiving unit, implemented by the control unit 46A of the robot 414, is used to receive user input. The parsing unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for parsing the context. The proposal unit, for example, is implemented by the specific processing unit 290 of the data processing device 12, for proposing proper nouns. The selection unit, for example, is implemented by the control unit 46A of the robot 414, for reflecting the user's selection. The correspondence between the various units and the device or control unit is not limited to the above examples and can be varied.

[0146] Furthermore, the emotion-specific model 59, serving as an emotion engine, can determine the user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine the user's emotion based on an emotion graph that serves as a specific mapping (see...). Figure 9 The system determines the user's emotions. In addition, the emotion-specific model 59 can also determine the robot's emotions in the same way, and the specific processing unit 290 can also perform specific processing using the robot's emotions.

[0147] Figure 9 This is a diagram representing an emotion map 400 that maps various emotions. In the emotion map 400, emotions are arranged radially from the center in concentric circles. The closer to the center of the concentric circles, the more primitive the emotion is. Further out on the concentric circles, emotions are arranged representing states or actions arising from mood. Emotion is a concept that includes both feelings and mental states. To the left of the concentric circles, emotions generated by reactions occurring in the brain are arranged roughly. To the right of the concentric circles, emotions guided by situational judgments are arranged roughly. Above and below the concentric circles, emotions generated by reactions occurring in the brain and guided by situational judgments are arranged roughly. Furthermore, the emotion of "pleasure" is arranged above the concentric circles, and the emotion of "unpleasantness" is arranged below. Thus, in the emotion map 400, various emotions are mapped according to the structure of emotion generation, while easily generated emotions are mapped nearby.

[0148] These emotions are distributed at the 3 o'clock position on the Emotion Chart 400, and usually fluctuate between peace and unease. In the right half of the Emotion Chart 400, because situational awareness is more dominant than internal feelings, it gives a sense of calm.

[0149] The inner side of the emotion diagram 400 represents the mind, and the outer side of the emotion diagram 400 represents actions. Therefore, the further you go to the outer side of the emotion diagram 400, the more the emotion can be seen (manifested in actions).

[0150] Here, human emotions are based on a balance of various factors such as posture and blood sugar levels. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. In robots, cars, and motorcycles, emotions can also be created based on a balance of factors such as posture and remaining battery power. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Speech Emotion Recognition and Brain Physiological Signal Analysis Systems for Emotion, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). In the left half of the emotion map, emotions belonging to the "Reaction" domain, where sensation is dominant, are arranged. Furthermore, in the right half of the emotion map, emotions belonging to the "Situation" domain, where situational cognition is dominant, are arranged.

[0151] The emotion map defines two types of emotions that promote learning. One is a negative emotion located near the middle of "repentance" or "reflection" on the situation side. That is, when the robot experiences negative emotions such as "I never want to feel this way again" or "I never want to be scolded again." The other is a positive emotion located near "desire" on the response side. That is, when the robot experiences positive feelings such as "wanting more" or "wanting to know more."

[0152] The emotion-specific model 59 feeds user input into a pre-learned neural network to obtain emotion values ​​representing each emotion shown in the emotion graph 400, and determines the user's emotion. This neural network is pre-learned based on multiple learning data sets that combine user input with emotion values ​​representing each emotion shown in the emotion graph 400. Furthermore, this neural network is learned to... Figure 10 As shown in sentiment graph 900, sentiment values ​​in nearby configurations are similar to each other. Figure 10 Examples show how emotions such as "peace of mind," "stability," and "reassurance" can result in similar emotional values.

[0153] In the above embodiments, a specific processing is described by a single computer 22, but the technology disclosed herein is not limited to this, and distributed processing by multiple computers, including computer 22, is also possible.

[0154] In the above embodiments, an example of storing a specific processing program 56 in memory 32 is illustrated, but the technology disclosed herein is not limited thereto. For example, the specific processing program 56 may also be stored in a portable computer-readable non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 performs specific processing according to the specific processing program 56.

[0155] Alternatively, the specific processing program 56 can be stored in a storage device such as a server connected to the data processing device 12 via a network 54, and the specific processing program 56 can be downloaded and installed into the computer 22 upon request from the data processing device 12.

[0156] Furthermore, it is not necessary to store the entire specific process 56 in a storage device such as a server connected to the data processing device 12 via the network 54, nor is it necessary to store the entire specific process 56 in the memory 32; a portion of the specific process 56 may also be stored.

[0157] As a hardware resource for performing specific processing, various processors can be used. For example, a CPU is a general-purpose processor that functions as a hardware resource for performing specific processing by executing software, i.e., programs. Additionally, a dedicated circuit can be listed as a processor; it is a processor with a circuit structure specifically designed for performing specific processing, such as a FPGA (Field-Programmable Gate Array), PLD (Programmable Logic Device), or ASIC (Application Specific Integrated Circuit). Every processor has built-in or connected memory, and every processor executes specific processing by using that memory.

[0158] The hardware resources for performing a specific process can consist of one of these various processors, or a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware resources for performing a specific process can also be a single processor.

[0159] As an example of a single processor, the first type consists of a combination of one or more CPUs and software, which functions as a hardware resource to perform specific processing. The second type uses a processor, such as a System-on-a-chip (SoC), which implements the entire system functionality, including multiple hardware resources performing specific processing, using a single IC chip. In this case, the specific processing is implemented using one or more of the aforementioned processors that serve as hardware resources.

[0160] Furthermore, as the hardware architecture of these various processors, more specifically, circuits combining semiconductor elements and other circuit components can be used. Moreover, the specific process described above is merely an example. Therefore, it goes without saying that, without departing from the main point, unnecessary steps can be removed, new steps can be added, or the processing order can be changed.

[0161] Furthermore, although the above examples have been described in terms of first to fourth embodiments, some or all of these embodiments can also be combined. Additionally, the smart device 14, smart glasses 214, head-mounted terminal 314, and robot 414 are just examples; they can be combined separately or are other devices.

[0162] The foregoing descriptions and illustrations are detailed explanations of the parts covered by this disclosure and are merely one example of this disclosure. For instance, the descriptions of the above-described structure, function, role, and effect are just one example of the structure, function, role, and effect of the parts covered by this disclosure. Therefore, it goes without saying that, without departing from the spirit of this disclosure, unnecessary parts can be deleted, new elements can be added, or replacements can be made to the foregoing descriptions and illustrations. Furthermore, to avoid confusion and facilitate understanding of the parts covered by this disclosure, explanations of technical common sense that does not require special explanation for implementing this disclosure have been omitted from the foregoing descriptions and illustrations.

[0163] All documents, patent applications and technical standards described in this specification are incorporated herein by reference as if they were specifically and individually described as incorporated by reference.

[0164] (Note 1) A system comprising: The reception department is used to receive user input. The parsing unit parses the input context received by the receiving unit; The proposal department proposes proper nouns based on the context analyzed by the analysis department. The selection section allows users to choose the proper nouns proposed by the proposal section.

[0165] (Note 2) The system as described in Appendix 1 is characterized in that, The parsing unit includes a reference unit for referencing past dialogue history or a general database.

[0166] (Note 3) The system as described in Appendix 1 is characterized in that, The proposal department includes a candidate suggestion department for suggesting multiple candidate options.

[0167] (Note 4) The system as described in Appendix 1 is characterized in that, The selection section includes a feedback section for reflecting the user's selection.

[0168] (Note 5) The system as described in Appendix 1 is characterized in that, The receiving department estimates the user's emotions and adjusts the timing of input assistance based on the estimated user emotions.

[0169] (Note 6) The system as described in Appendix 1 is characterized in that, The processing department analyzes the user's past input history and selects an appropriate processing method.

[0170] (Note 7) The system as described in Appendix 1 is characterized in that, The receiving department automatically begins processing requests based on the content of the conversation, using specific keywords or phrases as triggers.

[0171] (Note 8) The system as described in Appendix 1 is characterized in that, The receiving department estimates the user's emotions and determines the priority of accepting inputs based on the estimated user emotions.

[0172] (Note 9) The system as described in Appendix 1 is characterized in that, The receiving department prioritizes handling regionally specific proper nouns based on the user's geographical location information.

[0173] (Postscript 10) The system as described in Appendix 1 is characterized in that, The processing department analyzes users' social media activities and prioritizes processing related inputs.

[0174] (Postscript 11) The system as described in Appendix 1 is characterized in that, The parsing unit infers the user's emotions and adjusts the accuracy of context parsing based on the inferred user emotions.

[0175] (Postscript 12) The system as described in Appendix 1 is characterized in that, The parsing unit refers to past dialogue history to improve parsing accuracy during contextual analysis.

[0176] (Postscript 13) The system as described in Appendix 1 is characterized in that, The parsing unit applies different parsing algorithms based on the type of dialogue during context parsing.

[0177] (Postscript 14) The system as described in Appendix 1 is characterized in that, The analysis unit infers the user's emotions and adjusts the display method of the analysis results according to the inferred user emotions.

[0178] (Postscript 15) The system as described in Appendix 1 is characterized in that, The parsing unit performs contextual analysis based on the user's geographical location information.

[0179] (Postscript 16) The system as described in Appendix 1 is characterized in that, The parsing unit refers to relevant news or trend information to improve parsing accuracy during contextual analysis.

[0180] (Postscript 17) The system as described in Appendix 1 is characterized in that, The proposal department infers the user's emotions and adjusts the way the proposal is expressed based on the inferred user emotions.

[0181] (Postscript 18) The system as described in Appendix 1 is characterized in that, When submitting proposals, the proposal department adjusts the level of detail in the proposals based on the importance of proper nouns.

[0182] (Postscript 19) The system as described in Appendix 1 is characterized in that, When submitting proposals, the proposal department applies different proposal algorithms based on the category of proper nouns.

[0183] (Postscript 20) The system as described in Appendix 1 is characterized in that, The proposal department estimates the user's emotions and adjusts the length of the proposal based on the estimated user emotions.

[0184] (Postscript 21) The system as described in Appendix 1 is characterized in that, When submitting proposals, the proposal department determines the priority of proposals based on the timing of the submission of proper nouns.

[0185] (Postscript 22) The system as described in Appendix 1 is characterized in that, When submitting proposals, the proposal department adjusts the order of the proposals based on the relevance of proper nouns.

[0186] (Postscript 23) The system as described in Appendix 1 is characterized in that, The selection unit presupposes the user's emotions and adjusts the selection method based on the presumed user emotions.

[0187] (Postscript 24) The system as described in Appendix 1 is characterized in that, When making a selection, the selection unit refers to the user's past selection history to provide the optimal selection method.

[0188] (Postscript 25) The system as described in Appendix 1 is characterized in that, When making a selection, the selection section customizes the options based on the user's current conversation content.

[0189] (Postscript 26) The system as described in Appendix 1 is characterized in that, The selection unit presupposes the user's emotions and determines the priority of selection based on the presumed user emotions.

[0190] (Postscript 27) The system as described in Appendix 1 is characterized in that, When making a selection, the selection unit provides the optimal option based on the user's geographical location information.

[0191] (Postscript 28) The system as described in Appendix 1 is characterized in that, When making a selection, the selection unit analyzes the user's social media activity and proposes options.

[0192] (Postscript 29) The system as described in Appendix 2 is characterized in that, The reference unit presumes the user's emotions and selects reference data based on the presumed user emotions.

[0193] (Note 30) The system as described in Appendix 2 is characterized in that, When making references, the reference unit refers to past dialogue history to optimize the parsing algorithm.

[0194] (Postscript 31) The system as described in Appendix 2 is characterized in that, The reference unit estimates the user's emotions and adjusts the frequency of the reference based on the estimated user emotions.

[0195] (Note 32) The system as described in Appendix 2 is characterized in that, When referencing data, the reference unit weights the reference data according to the timing of the dialogue submission.

[0196] (Postscript 33) The system as described in Appendix 3 is characterized in that, The candidate prompting unit infers the user's emotions and adjusts the prompting method based on the inferred user emotions.

[0197] (Postscript 34) The system as described in Appendix 3 is characterized in that, When providing candidate suggestions, the candidate suggestion unit refers to the user's past selection history to suggest the best candidate option.

[0198] (Postscript 35) The system as described in Appendix 3 is characterized in that, The candidate suggestion unit infers the user's emotions and determines the priority of the candidates based on the inferred user emotions.

[0199] (Postscript 36) The system as described in Appendix 3 is characterized in that, When providing candidate suggestions, the candidate suggestion unit suggests the optimal candidate based on the user's geographical location information.

[0200] (Postscript 37) The system as described in Appendix 4 is characterized in that, The response unit estimates the user's emotions and adjusts the response method based on the estimated user emotions.

[0201] (Postscript 38) The system as described in Appendix 4 is characterized in that, When providing feedback, the feedback unit refers to the user's past selection history to provide the optimal feedback method.

[0202] (Postscript 39) The system as described in Appendix 4 is characterized in that, The response unit estimates the user's emotions and determines the priority of the response based on the estimated user emotions.

[0203] (Postscript 40) The system as described in Appendix 4 is characterized in that, When responding, the response unit provides the optimal response method based on the user's geographical location information.

Claims

1. A system, characterized in that, include: The reception department is used to receive user input. The parsing unit parses the input context received by the receiving unit; The proposal department proposes proper nouns based on the context analyzed by the analysis department. The selection section allows users to choose the proper nouns proposed by the proposal section.

2. The system as described in claim 1, characterized in that, The parsing unit includes a reference unit for referencing past dialogue history or a general database.

3. The system as described in claim 1, characterized in that, The proposal department includes a candidate suggestion department for suggesting multiple candidate options.

4. The system as described in claim 1, characterized in that, The selection section includes a feedback section for reflecting the user's selection.

5. The system as described in claim 1, characterized in that, The receiving department estimates the user's emotions and adjusts the timing of input assistance based on the estimated user emotions.

6. The system as described in claim 1, characterized in that, The processing department analyzes the user's past input history and selects an appropriate processing method.

7. The system as described in claim 1, characterized in that, The receiving department automatically begins processing requests based on the content of the conversation, using specific keywords or phrases as triggers.

8. The system as described in claim 1, characterized in that, The receiving department estimates the user's emotions and determines the priority of accepting inputs based on the estimated user emotions.

9. The system as described in claim 1, characterized in that, The receiving department prioritizes handling regionally specific proper nouns based on the user's geographical location information.

Images