Device configured to process user request by using language model or method therefor

Abstract

According to one embodiment of the present invention, proposed is a device configured to process a user request by using a language model. The device comprises: a memory configured to store a code for identifying user intention from the user request, executing a function corresponding to the user intention, and processing the user request; and a processor configured to execute the code for performing an operation for processing the user request, wherein the operation includes acquiring the user intention from the user request, selecting one of a plurality of preset functions corresponding to the user intention, processing the user request by using the selected function, and outputting the processed result of the user request to a display, and the acquiring of the user intention or the processing of the user request may be performed by using the language model.

Description

A device configured to process user requests using a language model or a method for the same

[0001] The present invention relates to a device configured to process user requests using a language model or a method for the same, and more specifically, to a user request processing device or a method for the same that provides various functions such as device control, searching for files or images, and translation according to an input user request.

[0002] Conversational AI language models are being commercialized, enabling services capable of generating human-like conversations based on natural language processing technology. Primarily based on the GPT (Generative Pre-trained Transformer) architecture, they are designed to understand and respond to various inputs like humans.

[0003] The key features of conversational AI language models lie in language comprehension and generation capabilities, such as understanding sentences input by users and generating natural responses based on them or responding to various languages ​​and topics, as well as large-scale pre-trained data that can be provided by possessing rich knowledge and contextual understanding abilities acquired through prior training with diverse text data available on the internet.

[0004] Conversational AI models are commonly classified as Large Language Models (LLMs); however, since these can be utilized through networks such as the Internet, expectations are rising for Small Language Models (SLMs) that can be provided on-device.

[0005] On-device SLM is a lightweight language model that runs on a device (e.g., PC, smartphone, IoT device, smart home appliance, etc.). It is a technology designed to perform natural language processing tasks on the device itself without relying on cloud servers or external data centers.

[0006] On-device SLMs utilize optimized model sizes considering the device's limited computational resources (CPU, GPU) and memory (RAM, storage). They are designed as lightweight models requiring relatively fewer parameters and operations compared to Large Language Models (LLMs) and are processed internally within the device without a network. As a result, sensitive user data is not transmitted externally, which can enhance privacy protection.

[0007] The present invention proposes a device configured to process user requests using an on-device SLM, or a method for the same.

[0008] The present invention proposes a device configured to process user requests or a method for doing so.

[0009] The problems to be solved by the present invention are not limited to the problems to be solved above, and other problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0010] According to one embodiment of the present invention, an apparatus configured to process a user request using a language model is proposed, wherein the apparatus comprises: a memory configured to store code for processing the user request by identifying a user intent in the user request and executing a function corresponding to the user intent; and a processor configured to execute the code for performing an operation to process the user request, wherein the operation includes obtaining the user intent in the user request, selecting one of a plurality of preset functions corresponding to the user intent, processing the user request using the selected function, and outputting the result of processing the user request to a display, wherein the obtaining of the user intent or the processing of the user request may be performed using the language model.

[0011] Additionally or alternatively, the operation of acquiring the user intent may include inputting a prompt to select one of the preset multiple functions along with the user request into the language model.

[0012] Additionally or alternatively, the above operation may include obtaining a function corresponding to the user intent from the language model in response to the input prompt.

[0013] Additionally or alternatively, the operation of processing the user request using the selected function may include entering a command prompt for processing the user request related to the selected function into the language model.

[0014] Additionally or alternatively, the operation of processing the user request using the selected function may include processing the user request in a function handler that processes the selected function.

[0015] Additionally or alternatively, the operation may include obtaining a processing result of the user request based on the user intent related to the selected function from the language model in response to the input command prompt.

[0016] Additionally or alternatively, the command prompt for processing the above user request may include a user intent related to the selected function.

[0017] Additionally or alternatively, the above operation may include outputting a user interface to the display for receiving the user request.

[0018] Additionally or alternatively, the user interface for receiving the above user request may include a first user interface for keyword-type user requests or a second user interface for conversational user requests.

[0019] Additionally or alternatively, the language model may be stored in the memory or the local storage space of the device.

[0020] Additionally or alternatively, the plurality of functions may include two or more of the following: control of a device based on the user request, summary of a document or text based on the user request, translation of a document or text based on the user request, search of a locally stored document based on the user request, search of the content of a locally stored document based on the user request, and search of a locally stored manual based on the user request.

[0021] Additionally or alternatively, the operation may include extracting keywords from the user request, matching the extracted keywords with pre-stored index information, and obtaining a processing result for the user request based on the matching.

[0022] Additionally or alternatively, the operation may include obtaining information about a device control, file, or image corresponding to a keyword of the user request using any one of the following: device control index information, which stores together text representing the control of the device and the execution path of each control; file index information, which stores together text extracted from a file stored locally on the device and meta information including at least the name, storage location, and modification date of the file; image index information, which stores together text extracted from an image stored locally on the device and meta information including at least the name, storage location, and modification date of the image; or file or image name index information, which stores together the name of the file or image stored locally and meta information including at least the name, storage location, and modification date of the file or image.

[0023] Additionally or alternatively, the operation may include outputting information about a device control, file, or image corresponding to the keyword of the acquired user request to the display.

[0024] Additionally or alternatively, the above operation may include outputting information regarding device control, a file, or an image in real time to the display in response to a user request in a state entered into a user interface for receiving the user request.

[0025] Additionally or alternatively, the device may include a human-machine interface for receiving the user request or outputting the result of processing the user request.

[0026] According to another embodiment of the present invention, a method for processing a user request using a language model is proposed, wherein the method is performed by a device comprising: a memory configured to store code for processing the user request by identifying a user intention in the user request and executing a function corresponding to the identified intention; and a processor configured to execute the code for performing an operation to process the user request, and comprises the steps of: obtaining the user intention in the user request; selecting one of a plurality of preset functions corresponding to the user intention; processing the user request using the selected function; and outputting the result of processing the user request to a display, wherein the obtaining of the user intention or the processing of the user request may be performed using the language model.

[0027] Additionally or alternatively, the above method may include the step of outputting a user interface to the display for receiving the user request.

[0028] Additionally or alternatively, the language model may be stored in the memory or the local storage space of the device.

[0029] According to another embodiment of the present invention, a computer-readable medium is proposed that stores code configured to be executed by a computer or processor for processing a user request using the aforementioned language model.

[0030] The above-mentioned problem-solving methods are merely some of the embodiments of the present invention, and various embodiments reflecting the technical features of the present invention can be derived and understood by those skilled in the art based on the detailed description of the present invention to be described below.

[0031] The present invention has the following effects.

[0032] The present invention can provide a conversational artificial intelligence model or a generative artificial intelligence language model locally without a network connection.

[0033] In addition, since the present invention processes user requests only locally, it has an advantage in terms of privacy protection.

[0034] In addition, the present invention can provide information or functions stored in the device to the user at high speed and with high accuracy.

[0035] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below.

[0036] The accompanying drawings, which are included as part of the detailed description to aid in understanding the present invention, provide embodiments of the present invention and explain the technical concept of the present invention together with the detailed description.

[0037] FIG. 1 illustrates a user interface for receiving a user request according to the present invention.

[0038] FIG. 2 illustrates a system structure diagram for processing user requests according to the present invention.

[0039] FIG. 3 illustrates various functions of an interactive model according to the present invention.

[0040] FIG. 4 illustrates a user request processing procedure using an interactive model according to the present invention.

[0041] FIG. 5 illustrates a procedure for determining user intent or selecting a function based on user intent using an interactive model according to the present invention.

[0042] FIG. 6 illustrates a user request processing procedure using a selected function using an interactive model according to the present invention.

[0043] FIG. 7 illustrates a user request processing procedure using a selected function using an interactive model according to the present invention.

[0044] FIG. 8 illustrates a user request processing procedure using a keyword-type model according to the present invention.

[0045] FIG. 9 illustrates various functions of a keyword-type model according to the present invention.

[0046] FIG. 10 illustrates an index information processor of a keyword-type model and various index information according to the present invention.

[0047] FIG. 11 illustrates the matching between keywords and index information obtained from a user request according to the present invention.

[0048] FIG. 12 illustrates a user interface that outputs the result of processing a user request according to the present invention.

[0049] FIG. 13 illustrates a user interface that outputs the result of processing a user request according to the present invention.

[0050] FIG. 14 illustrates a block diagram of a user request processing device using a language model according to the present invention.

[0051] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components regardless of drawing symbols will be assigned the same reference number, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not inherently possess distinct meanings or roles. Furthermore, in describing embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0052] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0053] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0054] A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0055] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0056]

[0057] FIG. 1 illustrates a user interface for receiving a user request according to the present invention.

[0058] The user interface (100) may include a first user interface component (10) for selecting either a keyword-type model or an interactive model.

[0059] Additionally, the user interface (100) may include a second user interface component (20) for displaying various functions provided by the user request processing device or method according to the present invention.

[0060] Additionally, the user interface (100) may include a third user interface component (30) for receiving user requests or prompts, which includes a fourth user interface component (31) for attaching files and a fifth user interface component (32) for ending or completing the input of user requests.

[0061] Additionally, the user interface (100) may include a sixth user interface component (40) for user input for refreshing.

[0062] Additionally, the user interface (100) can be output to a human-machine interface (HMI), such as a display.

[0063] The user interface (100) illustrated in FIG. 1 is merely an example, and the present invention may be implemented with a user interface (100) that includes other appearances or other user interface components.

[0064]

[0065] FIG. 2 illustrates a system structure diagram for processing user requests according to the present invention.

[0066] A system for processing user requests according to the present invention may include a user interface app (1000), a server app (2000), an interactive model (3100), and a keyword model (3200). All illustrated configurations (1000, 2000, 3100, 3200) are installed, mounted, embedded, or stored in a user request processing device (10) according to the present invention, and the present invention does not interact with other devices or systems outside the device (10). That is, the present invention performs user request processing locally on the device (10).

[0067] The device (10) of the present invention will be described later with reference to FIG. 14.

[0068] The user interface app (1000) is an app for interacting with a user, and is intended to output the user interface (100) shown in FIG. 1 through an HMI such as a display, receive user requests, and output the processing results for user requests.

[0069] The user interface app (1000) can store user requests entered in the past and the results of their processing, and can set or store user settings related to user request processing.

[0070] The server app (2000) is connected to the user interface app (1000) and is configured to transmit user requests entered through the user interface app (1000) to the conversational model (3100) or keyword model (3200), and to provide the processed results therefrom to the user interface app (1000).

[0071] The conversational model (3100) is a configuration for processing user requests entered in the form of a conversation using the previously described SLM, and will be described later with reference to FIGS. 3 to 7.

[0072] The keyword type model (3200) is a configuration for processing user requests entered in the form of keywords, and will be described later with reference to FIGS. 8 to 11.

[0073]

[0074] FIG. 3 illustrates various functions of an interactive model according to the present invention.

[0075] The conversational model (3100) consists of a language model (SLM) processor (3110) and various function processors (3111, 3112, 3113, 3114, 3115, 3116) that respond to user requests. The conversational model (3100) is embedded in or stored in the device (10) according to the present invention. Additionally, the conversational model may be controlled or executed by a processor of the device (10).

[0076] The language model processor (3110) receives a user request and transmits it to the function selector (3120), the language model (3130), or each function processor (3111, 3112, 3113, 3114, 3115, 3116), and can receive the result processed by the language model or each function processor.

[0077] There are a total of six illustrated function processors, but the present invention is not limited thereto. Functions not illustrated or described may be provided, and only some of the illustrated function processors may be provided.

[0078] The device control handler (3111) can handle functions for controlling the device (10). The control of the device (10) is related, for example, to enabling / disabling dark mode, enabling / disabling reader mode, or executing a specific app.

[0079] The summary processor (3112) can process the function of providing a summary of text entered as a user request or an entered / attached document.

[0080] The translation processor (3113) can process the function of providing a translation of text entered as a user request or an entered / attached document.

[0081] The document search handler (3114) can handle the search function for documents stored locally. The document search handler (3114) does not search the content of the document, but can support a search based on the metadata of the document (file) (e.g., file name, storage location, modification date, etc.).

[0082] The document content search handler (3115) can process the content search function for documents stored locally. Unlike the document search handler (3114), the document content search handler (3115) supports searching for the content of the document. To this end, the content of the document may be preprocessed in advance and text organized, keywords may be extracted or a summary may be generated to organize key information, or the document may be represented as a numerical vector based on the content of the document, i.e., key keywords. Additionally, indexing may be performed to efficiently search the numerical vector. Indexing may be performed by mapping words included in the document to the document (document name).

[0083] The manual search handler (3116) can perform a function to search for content of a manual stored locally. The manual may include information including user manuals of the device (10) or instructions related to the device (10), such as methods for dealing with failures.

[0084] Except for the device control processor (3111), all processing results of the user request can be expressed as information in text form. The device control processor (3111) can also provide results in text form, for example, information regarding device control corresponding to the user request in text form. Preferably, the device control processor (3111) can provide or perform the execution of device control corresponding to the user request and provide the processing results of the user request by outputting a notification message guiding the processing results to a user interface, etc.

[0085]

[0086] FIG. 4 illustrates a user request processing procedure using an interactive model according to the present invention. The interactive model (3100) is embedded or stored in a device (10) according to the present invention. Additionally, the interactive model may be controlled or executed by a processor of the device (10). Below, the procedure of FIG. 4 will be explained with a focus on the interactive model (3100).

[0087] The interactive model (3100) can receive user requests from the user interface (1000) (S410).

[0088] The interactive model (3100) determines a user intent from a user request and can select one of a plurality of preset functions provided by the interactive model based on the determined user intent (S420). The plurality of functions may include two or more of the previously described device control, summary, translation, document search, document content search, and manual search.

[0089] The interactive model (3100) can process user requests using selected functions (S430).

[0090] The interactive model (3100) can transmit the processing result of a user request to the user interface (1000) (S440).

[0091] The user interface (1000) can output the result of processing the user request (S450).

[0092] In FIG. 4, the processing of a user request using an interactive model according to the present invention was described in general terms, and each step will be described in detail below.

[0093]

[0094] FIG. 5 illustrates a procedure for determining user intent or selecting a function based on user intent using an interactive model according to the present invention. The interactive model (3100) is embedded or stored in a device (10) according to the present invention. Additionally, the interactive model may be controlled or executed by a processor of the device (10). FIG. 5 explains S420 of FIG. 4 in more detail.

[0095] The conversational model (3100) may include a language model handler (3110), a feature selector (3120), and a language model (3130).

[0096] The language model processor (3110) can receive an interactive user request (S411). The language model processor (3110) can extract necessary information from the user request and generate a request or command message in a form that can be transmitted to the function selector (3120) or the language model (3130), and can transmit the generated message to the function selector (3120) or the language model (3130). Additionally, the language model processor (3110) can receive a result processed by the function selector (3120) or the language model (3130).

[0097] Specifically, the language model processor (3110) can transmit a function selection request corresponding to a user request received by the function selector (3120) (S421).

[0098] The function selector (3120) can generate a function selection command prompt in response to a received function selection request (S422). The function selection command prompt may include a user request. The function selection command prompt may implicitly or explicitly include a request or command to determine the user's intent from the user request.

[0099] The function selector (3120) can transmit the generated function selection command prompt to the language model (3130) (S423).

[0100] The language model (3130) can determine the user's intent from the user request and select one of a plurality of functions based on the user's intent (S424).

[0101] The language model (3130) can transmit the selected function or the user intent that formed the basis of the selected function (i.e., the user intent for the selected function) to the function selector (3120) (S425).

[0102] Then, the feature selector (3120) can transmit the selected feature or the user intent for the selected feature to the language model handler (3110).

[0103] According to this procedure, the conversational model (3100) or the language model processor (3110) can obtain information about the user's intent or the function corresponding to the user's intent from the user request.

[0104]

[0105] FIG. 6 illustrates a user request processing procedure using a selected function using an interactive model according to the present invention. The interactive model (3100) is embedded or stored in a device (10) according to the present invention. Additionally, the interactive model may be controlled or executed by a processor of the device (10). FIG. 6 explains S430 of FIG. 4 in more detail.

[0106] The language model processor (3110) can obtain information about the user intent or the function corresponding to the user intent from the user request (S420).

[0107] Then, the language model processor (3110) can transmit a user request processing command to a selected function processor (311X) (S431). For example, in the language model (3130), if it is determined that the user request is an intention regarding a text translation function, the selected function processor may be a translation processor (3113). The user request processing command may include a user request or a user intention. To this end, the language model processor (3110) may store the relationship between the user request, the user intention obtained therefrom, or the selected function.

[0108] The function handler (311X) can generate a command prompt for processing user requests (S432). The command prompt for processing user requests may include a user request or a user intent.

[0109] Then, the function handler (311X) can pass the generated command prompt to the language model (3130) (S433).

[0110] The language model (3130) can process user requests according to a command prompt for processing user requests (S434). For example, if the command prompt for processing user requests was a text translation, the language model (3130) can obtain a translation of the text included in the command prompt.

[0111] Then, the language model (3130) can pass the result of processing the user request to the function handler (311X) (S435).

[0112] The function handler (311X) can transmit the processing result of the user request to the language model handler (3110) (S436).

[0113] According to this procedure, the conversational model (3100) or the language model processor (3110) can obtain the processing result of the user request by using the user intent or a function corresponding to the user intent.

[0114]

[0115] Meanwhile, the document content search function may involve additional procedures.

[0116] The document content search processor (3115) may perform additional procedures between receiving a user request processing command (S431) and creating a command prompt for processing the user request (S432). In the case of document content search, the document content search processor (3115) may search (parse) the content of a locally stored document in relation to the user request and obtain the part of the document content related to the user request. The document content search processor (3115) may create a command prompt for processing the user request based on the part related to the user request (S432).

[0117]

[0118] In addition, the device control processor (3111) may not perform all the procedures shown in FIG. 6. Below, this will be briefly explained with reference to the figure.

[0119]

[0120] FIG. 7 illustrates a user request processing procedure using a selected function using an interactive model according to the present invention. The interactive model (3100) is embedded or stored in a device (10) according to the present invention. Additionally, the interactive model may be controlled or executed by a processor of the device (10). FIG. 7 explains S430 of FIG. 4 in more detail.

[0121] FIG. 7 is for S430, which is performed when the function processor (311X) is a function processor capable of directly executing user requests, such as the device control processor (3111).

[0122] The language model processor (3110) can obtain information about the user intent or the function corresponding to the user intent from the user request (S420).

[0123] Then, the language model processor (3110) can transmit a user request processing command to a selected function processor (311X) (S431). For example, if the language model (3130) determines that the user request is an intention regarding a device control function, the selected function processor may be a device control processor (3111). The user request processing command may include a user request or a user intention. To this end, the language model processor (3110) may store the relationship between the user request, the user intention obtained therefrom, or the selected function.

[0124] The function handler (311X) can process user requests according to the user request processing command (S432).

[0125] Then, the function handler (311X) can pass the result of processing the user request to the language model handler (3110) (S436).

[0126]

[0127] FIG. 8 illustrates a user request processing procedure using a keyword-type model according to the present invention.

[0128] The keyword-type model (3200) is embedded or stored in the device (10) according to the present invention. Additionally, the keyword-type model may be controlled or executed by the processor of the device (10). Below, the procedure of FIG. 8 will be explained with the keyword-type model (3200) as the focus.

[0129] The keyword model (3200) can receive user requests (S810).

[0130] The keyword model (3200) can extract keywords from a user request (S820). Then, the keyword model (3200) matches the extracted keywords with index information and obtains a processing result for the user request based on the matching of the keywords and index information (S830).

[0131] The keyword model (3200) can transmit at least some of the index information matching the extracted keyword, which is the result of processing a user request, to the user interface (1000) (S840).

[0132] The user interface (1000) can output the result of processing the user request (S850).

[0133]

[0134] FIG. 9 illustrates various functions of a keyword-type model according to the present invention. The keyword-type model (3200) is embedded or stored in a device (10) according to the present invention. Additionally, the keyword-type model (3200) may be controlled or executed by a processor of the device (10).

[0135] The keyword-type model (3200) consists of an index information processor (3210) and various function processors (3211, 3212, 3213, 3214) that respond to user requests.

[0136] The index information processor (3210) receives a user request, extracts a keyword (1) from it, and then matches the keyword (1) with the index information (3220) through the keyword-index matcher (3223) so that the results processed by each function processor can be received.

[0137] There are a total of four illustrated function processors, but the present invention is not limited thereto. Functions not illustrated or described may be provided, and only some of the illustrated function processors may be provided.

[0138]

[0139] FIG. 10 illustrates an index information processor of a keyword-type model and various index information according to the present invention.

[0140] The keyword type model (3200) may include device control index information (3221), file index information (3222), image index information (3223), and file / image name index information (3224) managed by the index information processor (3210).

[0141] The device control index information (3221) may include information that stores text representing the control of the device (10) and the execution path (or user interface component such as an execution icon) of each control.

[0142] The file index information (3222) may include information that stores text extracted from a file stored locally in the device (10) and meta information including at least the name, storage location, and modification date of the file.

[0143] Image index information (3223) may include information that stores text extracted from an image stored locally in the device (10) and meta information including at least the name, storage location, and modification date of the image.

[0144] The file / image name index information (3224) may include information that stores the name of a file or image stored locally in the device (10) and at least the name of the file or image, storage location, and modification date of the file or image.

[0145] The following shows an example of file index information. The extracted text information may be at least one.

[0146] Text Info File Name Save Location Modified Date… USA, Travel, Vacation USA_2024.xxxD: / work2023-12-03 9:50AM… … … … … …

[0147]

[0148] FIG. 11 illustrates the matching between keywords and index information obtained from a user request according to the present invention.

[0149] An index information processor (3210) or a keyword-index matcher (3230) can extract a keyword (1) from a user request. Then, the index information processor (3210) or the keyword-index matcher (3230) can match the extracted keyword (1) with the index information (3220; 3221, 3222, 3223, 3224) described above. That is, the index information processor (3210) or the keyword-index matcher (3230) can identify device control index information, file index information, image index information, and file / image name index information that contain the keyword of the user request among the index information.

[0150] Then, the index information processor (3210) or keyword-index matcher (3230) can obtain device control, file, image, or file / image name, etc. corresponding to the keyword of the user request.

[0151]

[0152] FIG. 12 illustrates a user interface that outputs the result of processing a user request according to the present invention.

[0153] The user interface (110) may include a first user interface component (11) indicating that a search by keyword is selected.

[0154] Additionally, the user interface (110) may include a second user interface component (21) for displaying the processing result of a user request related to device control.

[0155] Additionally, the user interface (110) may include a third user interface component (22) for displaying the processing result of a user request related to a file or image index.

[0156] Additionally, the user interface (110) may include a fourth user interface component (23) for displaying the processing result of a user request related to a file or image name index.

[0157] Additionally, the user interface (110) may include a fifth user interface component (24) for displaying the processing result of a user request related to an image (or video) index.

[0158] Meanwhile, as illustrated in FIG. 12, when a user request is input into a user interface component (30), the result of the user request can be displayed in real time on the user interface (21, 22, 23, 24) of FIG. 12, even if a command or request indicating the termination or completion of the user request input is not input through the user interface component (32). FIG. 12 illustrates a state in which the Korean character “디” is input as a user request, and shows the result of performing the user processing procedure according to FIG. 8 with the input user request “디”.

[0159]

[0160] FIG. 13 illustrates a user interface that outputs the result of processing a user request according to the present invention.

[0161] The user interface (120) may include a first user interface component (12) indicating that a search is selected via dialogue.

[0162] Additionally, the user interface (110) may include a second user interface component (25) for displaying the processing result of a user request related to a manual search function.

[0163] Additionally, the user interface (110) may include a third user interface component (26) for displaying the processing result of a user request related to a text translation function.

[0164] Additionally, the user interface (110) may include a fourth user interface component (27) for displaying the processing result of a user request related to a manual search function.

[0165]

[0166] FIG. 14 illustrates a block diagram of a user request processing device using a language model according to the present invention.

[0167] A user request processing device (10) using a language model (hereinafter referred to as “device (10)”) may include a memory (100) configured to store code for processing the user request by identifying the user intent in the user request and executing a function corresponding to the user intent.

[0168] Additionally, the device (10) may include a processor (101) configured to execute code stored in memory (100) for performing operations to process user requests.

[0169] Additionally, the device (10) may further include a human-machine interface (102) for receiving a user request or outputting the result of processing the user request.

[0170] The processor (101) can perform an operation to obtain the user intent from the user request.

[0171] The processor (101) can perform the operation of selecting one of a plurality of preset functions corresponding to the user's intention.

[0172] The processor (101) can perform an operation to process the user request using the selected function.

[0173] Additionally, the processor (101) can perform the operation of outputting the processing result of the above user request to a display.

[0174] The acquisition of user intent or the processing of user requests can be performed using a language model stored in memory (100) or in another local storage space of the device (10).

[0175] Additionally, the processor (101) can perform the operation of inputting a command prompt to select one of a plurality of preset functions along with a user request to a language model. The processor (101) can perform the operation of obtaining a function corresponding to the user's intention from the language model in response to the input command prompt.

[0176] Additionally, the processor (101) may perform the action of entering a command prompt for processing a user request related to a function selected as a language model. The command prompt for processing the user request may include a user intent related to the selected function.

[0177] In response to this, the processor (101) may perform an operation to obtain a processing result of a user request related to a function selected from a language model in response to a command prompt for processing a user request. Additionally, the processor (101) may include an operation to obtain a processing result of a user request based on a user intent related to a function selected from a language model in response to a command prompt for processing a user request.

[0178] The processor (101) can perform an action to have a function handler that processes a selected function process handle a user request. This corresponds to processing in cases where the function handler can directly process the user request, such as device control (3111).

[0179] The processor (101) can perform the operation of outputting a user interface to a display for receiving user requests. The user interface for receiving user requests may include a first user interface for keyword-type user requests or a second user interface for conversational user requests.

[0180] Multiple functions may include two or more of the following: control of a device based on a user request, summarization of a document or text based on a user request, translation of a document or text based on a user request, search of locally stored documents based on a user request, search of the content of locally stored documents based on a user request, and search of locally stored manuals based on a user request.

[0181] The processor (101) can process user requests as follows using a keyword-type model (3200).

[0182] The processor (101) can perform an operation of extracting keywords from a user request. Then, the processor (101) can perform an operation of matching the extracted keywords with pre-stored index information. The processor (101) can perform an operation of obtaining a processing result for the user request based on the matching of keyword-index information.

[0183] More specifically, the processor (101) can perform an operation to obtain information about a device control, file, or image corresponding to a keyword of the user request using at least one of the following: device control index information, which stores together text representing the control of the device and the execution path of each control; file index information, which stores together text extracted from a file stored locally in the device and meta information including at least the name, storage location, and modification date of the file; image index information, which stores together text extracted from an image stored locally in the device and meta information including at least the name, storage location, and modification date of the image; or file or image name index information, which stores together the name of the file or image stored locally and meta information including at least the name, storage location, and modification date of the file or image.

[0184] Additionally, the processor (101) can perform the operation of outputting the processing result of a user request to a human-machine interface such as a display.

[0185] More specifically, the processor (101) can perform an operation of outputting information about device control, files, or images corresponding to keywords of the acquired user request to a human-machine interface such as a display. Additionally, or alternatively, the processor (101) can perform an operation of outputting information about device control, files, or images corresponding to a user request entered into a user interface for receiving user requests in real time to a human-machine interface such as a display.

[0186]

[0187] In the foregoing, a user request processing device (10) using a language model according to the present invention has been described with reference to FIG. 14. Even without referring to FIG. 14, the user request processing device (10) using a language model according to the present invention may perform the operation according to the present invention as described above in FIG. 1 to FIG. 13.

[0188]

[0189] In addition, as another aspect of the present invention, the operation of the above-described proposal or invention may be provided as code that can be implemented, practiced, or executed by a "computer" (a comprehensive concept including a system on chip (SoC) or (micro)processor, etc.), or as a computer-readable storage medium or computer program product that stores or contains said code, and the scope of the present invention may be extended to said code or as a computer-readable storage medium or computer program product that stores or contains said code.

[0190]

[0191] The detailed description of the preferred embodiments of the present invention disclosed above is provided to enable those skilled in the art to implement and practice the present invention. Although the present invention has been described with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the present invention as described in the following claims. Accordingly, the present invention is not intended to be limited to the embodiments shown herein, but to be given the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A device configured to process user requests using a language model, A memory configured to store code for processing the user request by identifying the user intent in the above user request and executing a function corresponding to the user intent; and It includes a processor configured to execute the code for performing an operation to process the above user request, and The above operation is: Obtaining the user intent from the above user request, Select one of a plurality of preset functions corresponding to the above user intention, and Process the above user request using the above-selected function, and, It includes outputting the processing result of the above user request to a display, A device in which the acquisition of the above user intent or the processing of the above user request is performed using the above language model.

2. In paragraph 1, the operation of acquiring the user intent is, A device comprising inputting a prompt to select one of a plurality of preset functions along with the user request using the above language model.

3. In paragraph 2, the above operation is, A device comprising obtaining a function corresponding to the user intent from the language model in response to the input prompt.

4. In paragraph 1, the operation of processing the user request using the selected function is, A device comprising inputting a command prompt for processing the user request related to the selected function using the above language model.

5. In paragraph 1, the operation of processing the user request using the selected function is, A device comprising processing the user request in a function handler that processes the selected function.

6. In paragraph 4, the above operation is, A device comprising obtaining a processing result of the user request based on the user intent related to the selected function from the language model in response to the input command prompt.

7. In paragraph 4, the command prompt for processing the user request comprises a user intent related to the selected function.

8. In paragraph 1, the above operation is, A device comprising outputting a user interface to the display for receiving the above user request.

9. In claim 8, the device comprising a user interface for receiving the user request, the user interface comprising a first user interface for keyword-type user requests or a second user interface for conversational user requests.

10. In paragraph 1, the device, wherein the language model is stored in the memory or the local storage space of the device.

11. In paragraph 1, the plurality of functions are, Control of the device based on the above user request, Summary of a document or text based on the above user request, Translation of a document or text based on the above user request, Search for locally stored documents based on the above user request, Searching the content of locally stored documents based on the above user request, A device comprising two or more of the searches of locally stored manuals based on the above user request.

12. In paragraph 1, the above operation is, Extract keywords from the above user request, and Match the above extracted keywords with the pre-stored index information, and A device comprising obtaining a processing result for the user request according to the above matching.

13. In paragraph 1, the above operation is, Device control index information storing text representing the control of the above device and the execution path of each control together, File index information that stores text extracted from a file stored locally in the device, along with metadata including at least the name, storage location, and modification date of the file. Image index information that stores together text extracted from an image stored locally in the device and meta information including at least the name, storage location, and modification date of the image, or A device comprising obtaining information about a device control, file, or image corresponding to a keyword of a user request by using at least one of the file or image name index information that stores together the name of the file or image stored locally and at least the file or image name, storage location, and modification date of the file or image.

14. In Paragraph 13, the above operation is, A device comprising outputting information about a device control, file, or image corresponding to the keyword of the user request obtained above to the display.

15. In Paragraph 13, the above operation is, A device comprising outputting information regarding device control, a file, or an image in real time to the display in response to a user request in a state entered into a user interface for receiving the above user request.

16. In paragraph 1, the device is, A device comprising a human-machine interface for receiving the above user request or outputting the processing result of the above user request.

17. A method for processing a user request using a language model, wherein the method is performed by a device comprising: a memory configured to store code for processing the user request by identifying a user intent in the user request and executing a function corresponding to the user intent; and a processor configured to execute the code for performing an operation to process the user request. A step of obtaining the user intent from the above user request; A step of selecting one of a plurality of preset functions corresponding to the above user intention; A step of processing the above user request using the above-selected function; and It includes the step of outputting the processing result of the above user request to a display, A method in which the acquisition of the user intent or the processing of the user request is performed using the language model.

18. In paragraph 17, the above method is, A method comprising the step of outputting a user interface to the display for receiving the above user request.

19. In paragraph 17, the method wherein the language model is stored in the memory or the local storage space of the device.

20. A computer-readable medium storing code configured to execute a method according to any one of paragraphs 17 through 19 by a computer or processor.

Images