AI usage support device, AI usage support method, and program

Abstract

To provide an apparatus, a method, and a program that assist users in using AI. [Solution] The AI ​​usage assistance device of the present invention comprises a morphological analysis unit that performs morphological analysis on text information representing commands to be executed by the AI; keyword master data that stores pre-registered words, processing attribute codes corresponding to processing attributes predetermined for each word, and scores predetermined for the processing attribute codes; a keyword extraction unit that extracts words obtained by morphological analysis from the words stored in the keyword master data as keywords; a score calculation unit that calculates a total score by summing up the scores for each processing attribute code; and a first determination unit that decides to proceed to execute the processing attribute corresponding to the processing attribute code with the highest total score calculated by the score calculation unit.

Description

[Technical Field]

[0001] The present invention relates to an AI usage support device, AI usage support method, and program for determining whether or not a command should be input to an AI such as generative AI (artificial intelligence), and the optimal AI to which a command should be input. [Background technology]

[0002] Generative AI has seen remarkable development in recent years. Examples of generative AI (hereafter simply referred to as "AI") include ChatGPT, Microsoft 365 Copilot, and Gemini, but new products are being developed every day. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6735392 [Non-patent literature]

[0004] [Non-Patent Document 1] https: / / zenn.dev / knowledgesense / articles / 2b6aa64f27ea89 (searched on October 7, 2020) Summary of the Invention [Problem to be solved by the invention]

[0005] These AIs are very useful, but each product has its own unique features. In addition, the functions provided free of charge are generally limited, and some functions require payment. In this case, the fee is determined by the number of tokens, which is the size of the instructions input to the AI.

[0006] Therefore, in order to obtain the most appropriate response from the AI ​​while minimizing costs, it is necessary to input commands to the most appropriate AI in concise terms, i.e., with a small number of tokens, depending on the content of the command. Also, some commands, such as desktop operations, can be executed without using AI, so for such commands, costs can be reduced by processing them locally within the PC so as not to use AI.

[0007] The present invention has been made in light of these circumstances, and aims to provide an AI usage support device, an AI usage support method, and a program that support users' use of AI by determining whether or not a command needs to be input to an AI, determining the optimal AI to input a command to, and generating commands with a small number of tokens when inputting commands to an AI. [Means for solving the problem]

[0008] In order to achieve the above object, the present invention takes the following measures.

[0009] The first aspect is an AI usage assistance device comprising: a morphological analysis unit that performs morphological analysis on text information representing commands to be executed by AI (Artificial Intelligence); keyword master data that stores pre-registered words, processing attribute codes corresponding to processing attributes predetermined for each word, and scores predetermined for the processing attribute codes; a keyword extraction unit that extracts words obtained by morphological analysis from the words stored in the keyword master data as keywords; a score calculation unit that obtains, from the keyword master data, the processing attribute codes predetermined for the words extracted as keywords and the scores predetermined for the processing attribute codes, and calculates a total score by summing up the scores for each processing attribute code; a first determination unit that determines to proceed to execute a processing attribute corresponding to the processing attribute code with the highest total score calculated by the score calculation unit; and a second determination unit that determines whether to use AI to execute the processing attribute determined to proceed by the first determination unit, and if it is determined to use AI, further determines the AI ​​to use.

[0010] The second aspect is an AI usage assistance device of the first aspect, which further includes processing attribute master data that stores processing attribute codes in association with AIs to be used for executing processing attributes corresponding to the processing attribute codes, and the second decision unit refers to the processing attribute master data to determine whether or not to use AI for executing the processing attribute determined to be subsequent by the first decision unit, and if it is determined that AI should be used, further determines the AI ​​to be used.

[0011] The third aspect is the AI ​​usage assistance device of the first aspect, further comprising an input creation unit that creates input to the AI ​​by abbreviating text information when the second decision unit determines that AI should be used.

[0012] The fourth aspect is the AI ​​usage assistance device of the first aspect, further comprising a text information conversion unit that, when a command is input by voice, converts the voice into text information and provides the text information to the morphological analysis unit for morphological analysis.

[0013] The fifth aspect is an AI usage assistance method implemented by an AI usage assistance device, in which a processor of the AI ​​usage assistance device executes the following steps: a first step of morphologically analyzing text information representing commands to be executed by AI (Artificial Intelligence), a second step of extracting words obtained by the morphological analysis as keywords from among words prepared in advance, a third step of calculating a total score by summing up the scores for each processing attribute code based on processing attribute codes predetermined for the words extracted as keywords and scores predetermined for the processing attribute codes, a fourth step of deciding to proceed to execute a processing attribute corresponding to the processing attribute code with the highest total score, and a fifth step of deciding whether to use AI to execute the processing attribute decided to proceed, and if it is decided to use AI, further deciding the AI ​​to use.

[0014] The sixth aspect is a method for supporting the use of AI of the fifth aspect, in which the fifth step refers to the processing attribute master data stored in association with the processing attribute code and the AI ​​used to execute the processing attribute corresponding to the processing attribute code to determine whether or not to use AI to execute the processing attribute that has been determined to follow, and if it is determined that AI will be used, the method further determines the AI ​​to use.

[0015] A seventh aspect is a method for assisting in the use of AI according to the fifth aspect, in which the processor further performs a sixth step of creating input for the AI ​​by abbreviating the text information if it is determined in the fifth step that AI will be used.

[0016] An eighth aspect is the AI ​​usage assistance method of the fifth aspect, in which the processor further executes a seventh step of converting the voice into text information when the command is input by voice, and the first step is morphological analysis of the text information converted in the fifth step.

[0017] The ninth aspect is a program for causing a processor to realize the following functions: a function for morphologically analyzing text information representing commands to be executed by AI (Artificial Intelligence); a function for extracting words obtained by the morphological analysis as keywords from among words prepared in advance; a function for calculating a total score by summing up the scores for each processing attribute code based on the processing attribute code predetermined for the word extracted as a keyword and the scores predetermined for the processing attribute code; a function for deciding to proceed to the execution of a processing attribute corresponding to the processing attribute code with the highest total score; and a function for deciding whether to use AI in the execution of the processing attribute determined to proceed, and if it is decided to use AI, further deciding the AI ​​to use.

[0018] A tenth aspect is a program of the ninth aspect, in which the function of determining the AI ​​to be used refers to processing attribute master data in which the processing attribute code and the AI ​​to be used for executing the processing attribute corresponding to the processing attribute code are stored in association with each other, to determine whether or not to use an AI to execute the processing attribute that has been determined to be subsequent, and if it is determined that an AI will be used, further includes a function of determining the AI ​​to be used.

[0019] An eleventh aspect is a program of the ninth aspect, which, when it is determined that AI is to be used, further causes the processor to realize a function of creating input to the AI ​​by abbreviating the text information.

[0020] A twelfth aspect is the program of the ninth aspect, further causing the processor to implement a function of converting speech to text information to generate text information that is morphologically analyzed when the command is input by voice. [Effects of the Invention]

[0021] The AI ​​usage support device, method, and program of the present invention can determine whether or not to input commands to the AI, thereby reducing costs by executing commands on a local PC without using the AI.

[0022] The AI ​​usage support device, method, and program of the present invention also make it possible to determine the optimal AI to which a command should be issued, allowing the user to use an appropriate AI in response to the command.

[0023] Furthermore, the AI ​​usage support device, method, and program of the present invention can generate commands with fewer tokens when inputting commands to an AI, allowing users to use AI while minimizing costs associated with using AI. [Brief explanation of the drawings]

[0024] [Figure 1] FIG. 1 is a simplified block diagram illustrating the concept of the functional configuration of a typical AI assistance device. [Figure 2] FIG. 2 is a simplified block diagram illustrating the concept of the functional configuration of an AI usage assistance device to which a method for assisting use of AI according to an embodiment of the present invention is applied. [Figure 3] FIG. 3 is an electronic circuit block diagram showing an example of the configuration of an AI assistance device according to an embodiment of the present invention. [Figure 4] FIG. 4 is a functional block diagram corresponding to the electronic circuitry of the AI ​​assistance device illustrated in FIG. [Figure 5] FIG. 5 is a data structure diagram showing an example of keyword master data. [Figure 6] FIG. 6 is a data structure diagram showing an example of processing attribute master data. [Figure 7] FIG. 7 is a table for explaining the processing attributes. [Figure 8] FIG. 8 is a diagram showing the flow of processing by an AI usage assistance device to which the AI ​​usage assistance method according to the embodiment of the present invention is applied. [Figure 9] FIG. 9 is a diagram showing an example of abbreviations made by the AI ​​input creation unit. [Figure 10] FIG. 10 is a diagram for explaining the local processing. DETAILED DESCRIPTION OF THE INVENTION

[0025] Hereinafter, embodiments of the present invention will be described with reference to the drawings. The drawings are schematic or conceptual. In this specification and each drawing, elements similar to those described in the previous drawings are designated by the same reference numerals, and detailed and overlapping descriptions will be omitted as appropriate.

[0026] First, an overview of an AI usage assistance device to which a method for assisting the use of AI according to an embodiment of the present invention is applied will be described.

[0027] FIG. 1 is a simplified block diagram illustrating the concept of the functional configuration of a typical AI assistance device.

[0028] (1) A user U using a general AI usage assistance device 100 illustrated in FIG. 1 uses a microphone MC to input commands that the AI ​​wants to execute into a PC by voice.

[0029] (2) The PC converts this voice input into text information as commands for the AI. Specifically, the PC (2-1) converts the voice input into text information, and (2-3) generates a script for the AI. Hereinafter, as an example, the AI ​​usage assistance device 100 will be described as using ChatGPT as the AI.

[0030] (3) The PC inputs the ChatGPT script generated in (2-3) into ChatGPT. ChatGPT performs an analysis based on the ChatGPT script, and the analysis response is returned to the PC. For example, if the analysis result is a (3-1) screen operation command, the PC inputs the (4-1) screen operation command into the screen operation program as (4) feedback for the operation, and performs (4-2) PC screen operations according to the execution result of the (4-2) screen operation program.

[0031] As such, a fixed AI such as ChatGPT is used in a typical AI usage assistance device 100. In contrast, a device to which the AI ​​usage assistance method according to an embodiment of the present invention is applied has additional functions that a typical AI usage assistance device 100 does not have, as will be explained below.

[0032] FIG. 2 is a simplified block diagram illustrating the concept of the functional configuration of an AI usage assistance device to which a method for assisting use of AI according to an embodiment of the present invention is applied.

[0033] The AI ​​usage assistance device 10 includes a subsystem (2-2) that the AI ​​usage assistance device 100 does not include. Furthermore, the AI ​​used is not fixed; it can use any AI currently available, enabling it to accommodate not only the creation of new services but also the discontinuation of services. Therefore, the AI ​​usage assistance device 10 can use the subsystem (2-2) to (a) determine which AI to use from among multiple AIs, such as ChatGPT, Microsoft 365 Copilot, and Gemnini, based on the content of the command, and (b) generate and input a script to the selected AI. In this case, the number of tokens can be reduced by generating the script with as short a statement as possible. It can also (c) determine that there is no appropriate AI to use based on the content of the command. In this case, the device does not generate a script for the AI, i.e., it does not perform the operation (2-3), and does not input the script to the AI ​​(3). Instead, it generates screen operation commands (3-1) through local processing within the PC.

[0034] The specific configuration of the AI ​​usage assistance device 10 that realizes these functions will be described below.

[0035] FIG. 3 is an electronic circuit block diagram showing an example of the configuration of an AI assistance device according to an embodiment of the present invention.

[0036] FIG. 4 is a functional block diagram corresponding to the electronic circuitry of the AI ​​assistance device illustrated in FIG.

[0037] An AI usage assistance device 10 according to an embodiment of the present invention is a device for determining whether or not to input commands to an AI, such as a generative AI, determining the optimal AI to input commands to, and generating commands with a small number of tokens when inputting commands to an AI. As shown in Figure 3, the device includes a CPU 12, a recording medium reader 14, a display 16, a voice input unit 17 including a microphone, a text information input unit 18 including a mouse and keyboard, memory 20, a storage device 40, and a communications unit 50, all connected via a bus 11. Such an AI usage assistance device 10 can be implemented on, but is not limited to, a PC or a tablet device.

[0038] The CPU 12 is a computer that controls the operation of each circuit component in accordance with programs stored in the memory 20. This allows the AI-assisted device 10 to operate in a coordinated manner between software and hardware, as will be described later.

[0039] The memory 20 stores programs that realize a text information conversion unit 21, a morphological analysis unit 22, a keyword extraction unit 23, a score calculation unit 24, a first determination unit 25, a second determination unit 26, an AI input creation unit 27, an AI input input unit 28, and a local input input unit 29.

[0040] These programs may be stored in advance in memory 20, or may be read from an external recording medium 13 such as a memory card via recording medium reading unit 14 and stored in memory 20. These programs are designed not to be rewritable.

[0041] In addition to the non-rewritable area in which such programs are stored, the memory 20 also has a writable data area 30 secured as an area for storing rewritable data.

[0042] The storage device 40 is, for example, a solid state drive (SSD) or a hard disk drive (HDD), and stores keyword master data 42 and processing attribute master data 44.

[0043] FIG. 5 is a data structure diagram showing an example of keyword master data.

[0044] As illustrated in FIG. 5, the keyword master data 42 is a table that stores data corresponding to items consisting of a pre-registered word a, a processing attribute code b that is predetermined for each word a, a processing attribute name c that corresponds to the processing attribute code b, and a predetermined processing attribute score d.

[0045] FIG. 6 is a data structure diagram showing an example of processing attribute master data.

[0046] As illustrated in Figure 6, the processing attribute master data 44 is a table that stores data corresponding to items consisting of a processing attribute code b, a processing attribute name c corresponding to the processing attribute code b, and an "AI to use" e corresponding to the processing attribute code b.

[0047] Next, the processing attributes will be described.

[0048] FIG. 7 is a table for explaining the processing attributes.

[0049] The processing attribute classifies and defines the operations that the user wants to perform, and as shown by the processing attribute name c in Figure 7, can include, but is not limited to, PowerAutomate (RPA) operations, MS app operations, screen acquisition instruction operations, desktop operations, browser operations, and resolution processes. The processing attribute code b is the code number 01, 02, 03, 04, 05, or 80 assigned to these operations. An overview of each of these operations is as described in the processing attribute overview f in Figure 7.

[0050] The communication unit 50 can communicate with the AI ​​engine 80 via a communication network 60 such as the Internet. Therefore, the AI ​​usage assistance device 10 can send and receive data to and from the AI ​​engine 80 via the communication unit 50.

[0051] The AI ​​usage support device 10 determines whether or not a command needs to be input to an AI, determines the optimal AI to which the command should be input, and generates commands with a small number of tokens when inputting commands to an AI, based on text information.

[0052] The text information input unit 18 is a section where the user inputs text information for this purpose. The user inputs the content that the user wants to implement using AI as text information A from the text information input unit 18, which may include, for example, a mouse, a keyboard, etc.

[0053] Meanwhile, with the AI ​​assistance device 10, the user can input text information by voice using, for example, a microphone, rather than using a mouse or keyboard. Voice B input from the voice input unit 17 is converted into text information C by the text information conversion unit 21 using a technique such as that disclosed in Patent Document 1. This allows users who have difficulty using a mouse or keyboard due to a disability in their hands, users who cannot use their hands while working at a construction site, or users who cannot use their hands while cooking, to input necessary information into the AI ​​assistance device 10.

[0054] The morphological analysis unit 22 performs morphological analysis on the text information A from the text information input unit 18 and the text information C from the text information conversion unit 21, and outputs the result D to the keyword extraction unit .

[0055] The keyword extraction unit 23 refers to the keyword master data 42 and extracts, as keywords, words obtained as a result D of the morphological analysis by the morphological analysis unit 22 from words a pre-registered in the keyword master data 42.

[0056] The score calculation unit 24 acquires the processing attribute code b and score d that are pre-specified for the word a extracted as a keyword from the keyword master data 42. Then, the score calculation unit 24 calculates the total score E by summing up the scores d for each processing attribute code b.

[0057] Based on the total score E for each process attribute code b calculated by the score calculation unit 24, the first determination unit 25 determines to proceed to the execution of the process attribute corresponding to the process attribute code b with the highest total score.

[0058] The second decision unit 26 refers to the processing attribute master data 44 to determine whether or not to use AI to execute the processing attribute that has been determined to follow by the first decision unit 25, and if it is determined that AI will be used, it further determines the AI ​​to use.

[0059] When the second decision unit 26 determines that the command should be input to the AI, the AI ​​input creation unit 27 creates an input F for the AI ​​by abbreviating the text information A or the text information C.

[0060] The AI ​​input input unit 28 associates the input F created by the AI ​​input creation unit 27 with the optimal AI determined by the second determination unit 26 and outputs it to the communication unit 50.

[0061] The communication unit 50 transmits the input F via the communication network 60 to the AI ​​engine 80, which executes the optimal AI output from the AI ​​input input unit 28.

[0062] The AI ​​engine 80 executes processing based on the input F, and returns the processing result G to the communication unit 50 via the communication network 60. The user can view and confirm the processing result G on the display 16.

[0063] On the other hand, if the second decision unit 26 determines that the command should not be input to the AI, the local input input unit 29 creates command H for the local PC 70 based on the text information A or text information C, and causes the local PC 70 to execute the command. The local PC 70 is a PC on which the AI ​​usage assistance device 10 is implemented. Command H is also created by abbreviating the text information A or text information C.

[0064] The local PC 70 performs processing based on the command H. The user can view and confirm the processing result I on the display 16.

[0065] Next, an example of the operation of the AI ​​usage assistance device to which the AI ​​usage assistance method according to the embodiment of the present invention configured as described above is applied will be described.

[0066] FIG. 8 is a diagram showing the flow of processing by an AI usage assistance device to which the AI ​​usage assistance method according to the embodiment of the present invention is applied.

[0067] For example, as illustrated in the second row of Figure 8, we will explain the case where the user speaks the command they want the AI ​​to execute using the voice input unit 17, saying, "Sort column C in Excel in descending order."

[0068] The utterance content g of the voice B input from the voice input unit 17 is converted by the text information conversion unit 21 into text information C saying "sort column C in Excel in descending order."

[0069] This text information C is subjected to Japanese morphological analysis by the morphological analysis unit 22, and morphemes such as "Excel," "C," "column," "descending order," and "sort" are obtained. See the second line of "Japanese morphological analysis" h in FIG. 8.

[0070] The keyword extraction unit 23 refers to the keyword master data 42 and extracts, as keywords, words obtained by the morphological analysis unit 22 from the words a stored in the keyword master data 42. The keyword master data 42 shown in Fig. 5 stores words a such as "Excel," "cursor," "move," "screen," and "click," and among these, "Excel" is a word obtained by the morphological analysis unit 22, so in this case, the word a "Excel" is extracted as a keyword.

[0071] The score calculation unit 24 obtains from the keyword master data 42 the predetermined processing attribute code b "02" for the word a "Excel" extracted as a keyword, and the predetermined score d "1 point." Furthermore, the scores d for each processing attribute code b "02" are summed up to calculate the total score E "1 point." See the second line of "Calculate processing attribute score from keyword" i in Figure 8.

[0072] The first determination unit 25 determines to proceed to the execution of the processing attribute corresponding to the processing attribute code b having the highest total score E calculated by the score calculation unit 24. In this example, since the processing attribute code b is only "02", the processing attribute name of the processing attribute to be followed is "MS application operation". See the second line of "Subsequent processing branches" j in Figure 8.

[0073] The second decision unit 26 refers to the processing attribute master data 44 and determines whether to use AI for the processing attribute "MS app operation" determined to be subsequent by the first decision unit 25, and if it is determined that AI should be used, the AI ​​to be used is further determined.

[0074] The process attribute master data 44 specifies that the AI ​​to be used for "MS app operation," whose process attribute code b is "02," is "Copilot." Therefore, in this case, the second determination unit 26 determines that the AI ​​"Copilot" will be used. See the second line of "AI to be used in subsequent processing or not to be used" k in FIG. 8.

[0075] When the second determination unit 26 determines that the command should be input to the AI, the AI ​​input creation unit 27 creates an input F for the AI. At this time, if the text information A or text information C indicating the command is long, the input F is created by shortening the text information A or text information C.

[0076] FIG. 9 is a diagram showing an example of abbreviations made by the AI ​​input creation unit.

[0077] For example, as shown in Fig. 9, the AI ​​input creation unit 27 can shorten text information A consisting of 142 characters such as "Um, um, I have a favor to ask. I'd like you to teach me how to do conditional branching in Excel. Specifically, could you give me an example of the numbers in cells A1 and B1? And, um, I'd like you to set a function in cell C1, but I don't know how. Can you do that?" to create input F consisting of 53 characters such as "Teach me how to do conditional branching in Excel. Give me an example of setting a function in cell C1 using the numbers in cells A1 and B1." In this case, the text is shortened by about one-third.

[0078] For such abbreviated text, the AI ​​input creation unit 27 can use existing character compression technology such as that disclosed in Non-Patent Document 1.

[0079] The AI ​​input input unit 28 associates the input F created by the AI ​​input creation unit 27 with the optimal AI determined by the second determination unit 26 and outputs it to the communication unit 50.

[0080] The communication unit 50 transmits the input F via the communication network 60 to the AI ​​engine 80 that executes the AI ​​associated with the input F output from the AI ​​input input unit 28.

[0081] The AI ​​engine 80 executes processing based on the input F, and the processing result is returned from the AI ​​engine 80 to the communication unit 50 via the communication network 60. The user can view and confirm the processing result G on the display 16.

[0082] As mentioned above, the input F to the AI ​​is shortened, which reduces the amount of tokens required and reduces the cost of using the AI. Furthermore, shortening the input F also reduces the waiting time for a response from the AI, making it possible to improve response speed.

[0083] On the other hand, if the second decision unit 26 determines that the command should not be input to the AI, the local input input unit 29 creates a command H to be executed by the local PC 70 based on the text information A or the text information C, and inputs it to the local PC 70.

[0084] The local PC 70 performs processing based on the command H. The user can view and confirm the processing result I on the display 16.

[0085] FIG. 10 is a diagram for explaining such local processing.

[0086] For example, suppose that the user inputs text information A, "move cursor left," through the text information input unit 18. In this case, the morpheme analysis unit 22 obtains the morphemes "cursor," "left," and "move."

[0087] The keyword extraction unit 23 refers to the keyword master data 42 and extracts, as keywords, the words obtained by the morphological analysis unit 22 from the words a stored in the keyword master data 42. Of the words a stored in the keyword master data 42 shown in Fig. 5, the words a "cursor" and "move" are extracted as keywords.

[0088] The score calculation unit 24 obtains from the keyword master data 42 the processing attribute code b "04" predetermined for the words a extracted as keywords, "cursor" and "move," and the pre-specified score d "1 point." Furthermore, the scores d for each processing attribute code b "04" are summed up to calculate the total score E "2 points." See the third line of "Calculate processing attribute score from keywords" i in Figure 8.

[0089] The first determination unit 25 determines to proceed to the execution of the processing attribute corresponding to the processing attribute code b having the highest total score E calculated by the score calculation unit 24. In this example, since the processing attribute code b is only "04", the processing attribute name of the processing attribute to be followed is "Desktop Operation". See the third line of "Subsequent Processing Branches" j in Figure 8.

[0090] The second determination unit 26 refers to the process attribute master data 44 and determines whether or not to use AI for the process attribute "desktop operation" determined by the first determination unit 25 to be subsequent.

[0091] The processing attribute master data 44 specifies that AI is not used for "desktop operation" for which the processing attribute code b is "04." Therefore, in this case, the second determination unit 26 determines that AI is not used. See the third line of "AI to be used in subsequent processing or not to be used" k in Figure 8.

[0092] If the second decision unit 26 determines not to use AI, the local input input unit 29 creates a command H for the local PC 70. Like the AI ​​input creation unit 27, the local input input unit 29 can also create the command H by abbreviating the text information C indicating the command.

[0093] In response to this, the local PC 70 executes a desktop operation to move the cursor to the left based on the command H "move cursor to the left."

[0094] As described above, the AI ​​usage support device to which the AI ​​usage support method according to the embodiment of the present invention is applied can determine whether or not a command needs to be input to the AI ​​through the above-mentioned operation. As a result, if a command can be executed without using the AI, the user is notified of this fact, and the user can execute the command on their local PC without using the AI, thereby reducing costs.

[0095] The AI ​​usage assistance device to which the AI ​​usage assistance method according to an embodiment of the present invention is applied also prevents vendor lock-in and makes it possible to flexibly determine the optimal AI to which commands should be input from among any AIs by excluding AIs whose services have been discontinued from the list of candidates and adding AIs that have just started services as new candidates.

[0096] This allows users to always use the appropriate AI according to their commands, regardless of whether a service is discontinued or a new service is launched, for example by selecting and using the most high-performance AI at the lowest cost.

[0097] Furthermore, the AI ​​usage support device to which the AI ​​usage support method according to the embodiment of the present invention is applied can generate commands with a small number of tokens when issuing commands to the AI. This allows users to use the AI ​​while minimizing costs associated with using the AI. This also reduces the AI's response latency, leading to a significant improvement in the system's response speed.

[0098] Although the best mode for carrying out the present invention has been described above with reference to the accompanying drawings, the present invention is not limited to such a configuration. Those skilled in the art may conceive of various modifications and alterations within the scope of the technical ideas of the invention as defined in the claims, and it is understood that such modifications and alterations also fall within the technical scope of the present invention. [Explanation of symbols]

[0099] 10 AI usage support device 11 Bus 12 CPU 13 External storage media 14 Recording medium reading unit 16 Display 17 Audio input section 18 Text information input section 20 memory 21 Text information conversion unit 22 Morphological analysis section 23 Keyword Extraction Section 24 Point calculation section 25 First Decision Section 26 Second Decision Section 27 AI Input Creation Department 28 AI input section 29 Local input input section 30 Writable Data Area 40 Storage device 42 Keyword Master Data 44 Processing Attribute Master Data 50 Communications Department 60 Communication Network 70 local PC 80 AI Engine 100 AI usage support equipment a word A. Text information b Processing attribute code B Audio c Processing attribute name C. Text information d score D Morphological analysis results E Total points f Processing Attributes Overview F input g Vocal content G Processing results H command I Processing results U User

Claims

1. an input unit that accepts input of voice or text information representing a command to be executed by AI (Artificial Intelligence); keyword master data storing pre-registered words, processing attribute codes corresponding to processing attributes predetermined for each of the words, and scores predetermined for the processing attribute codes in association with each other; a keyword extraction unit that extracts, from among words stored in the keyword master data, words contained in the voice or text information input to the input unit as keywords; a score calculation unit that acquires, from the keyword master data, the processing attribute codes that are predetermined for the words extracted as the keywords and the scores that are predetermined for the processing attribute codes, and calculates a total score by summing up the scores for each of the processing attribute codes; a first determination unit that determines to execute an operation defined by a processing attribute corresponding to a processing attribute code having a high total score calculated by the score calculation unit; a second decision unit that determines whether or not to use AI to execute the operation determined to be executed by the first decision unit, and when it is determined that AI is to be used, further determines AI to be used; An AI usage support device equipped with the above.

2. an input unit that accepts input of voice or text information representing a command to be executed by AI (Artificial Intelligence); a keyword extraction unit that refers to keyword master data that associates pre-registered words with processing attribute codes corresponding to processing attributes predetermined for each of the words and scores predetermined for the processing attribute codes, and extracts, as keywords, words included in the speech or text information input to the input unit from among the words included in the keyword master data; a score calculation unit that acquires, from the keyword master data, the processing attribute codes that are predetermined for the words extracted as the keywords and the scores that are predetermined for the processing attribute codes, and calculates a total score by summing up the scores for each of the processing attribute codes; a first determination unit that determines to execute an operation defined by a processing attribute corresponding to a processing attribute code having a high total score calculated by the score calculation unit; a second decision unit that determines whether or not to use AI to execute the operation determined to be executed by the first decision unit, and when it is determined that AI is to be used, further determines AI to be used; An AI usage support device equipped with the above.

3. The method further includes processing attribute master data storing the processing attribute code and an AI used for executing an operation defined by the processing attribute corresponding to the processing attribute code in association with each other, The AI ​​usage support device according to claim 1 or 2, wherein the second decision unit refers to the processing attribute master data to determine whether to use AI to execute the operation determined to be executed by the first decision unit, and if it is determined that AI will be used, further determines the AI ​​to use.

4. 3. The AI ​​usage support device according to claim 1, further comprising an input creation unit that creates input to the AI ​​by shortening the command when the second decision unit determines that AI should be used.

5. An AI usage support method implemented by an AI usage support device, The processor of the AI ​​usage support device A first step of receiving input of voice or text information representing a command to be executed by AI (Artificial Intelligence); a second step of extracting words contained in the speech or text information as keywords from words prepared in advance; a third step of calculating a total score by summing up the scores for each processing attribute code based on processing attribute codes predetermined for the words extracted as the keywords and scores predetermined for the processing attribute codes; a fourth step of determining to execute an operation defined by a processing attribute corresponding to the processing attribute code with the highest total score; a fifth step of determining whether or not to use AI to execute the operation determined to be executed, and if it is determined that AI is to be used, determining the AI ​​to be used; A method for supporting the use of AI.

Images