Announcement sentence generation program, announcement sentence generation system, and announcement sentence generation method

The system efficiently addresses the inefficiencies in announcement generation by using a structured approach with large-scale language models and predefined examples to produce accurate and relevant announcements with minimal user input.

JP2025179540APending Publication Date: 2025-12-10TOA CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2024086369
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2024-05-28
Publication Date
2025-12-10

AI Technical Summary

Technical Problem

Existing technologies for generating announcements often require users to engage in trial and error when using trained models, leading to inefficiencies and time wastage in obtaining desired information.

Method used

An announcement generation system and method that includes a series of computational steps to acquire and process announcement source information, generate prompt information, and input it into a language processing computer to produce announcements efficiently, utilizing large-scale language models and predefined system example information to guide the generation process.

Benefits of technology

Enables the generation of announcements with a reduced number of trial and error attempts, ensuring accuracy and relevance to the user's input, while adhering to specified characteristics such as character count and avoiding prohibited terms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2025179540000001_ABST
    Figure 2025179540000001_ABST
Patent Text Reader

Abstract

To provide an announcement sentence generation program, an announcement sentence generation system, and an announcement sentence generation method capable of acquiring an announcement sentence with a small number of times of trial and error.SOLUTION: An announcement sentence generation program is a program for generating an announcement sentence. The announcement sentence generation program includes a first step to a third step. In the first step, a computer is caused to acquire announcement source information A1 for creating an announcement sentence. In the second step, the program causes the computer to form prompt information P1 for causing a language-processable computer to generate the announcement sentence, based on the announcement source information A1. In the third step, the program causes the computer to input the prompt information P1 formed in the second step to the language-processable computer, and acquire an announcement sentence from the language-processable computer.SELECTED DRAWING: Figure 3
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] The present disclosure relates to an announcement generation program, an announcement generation system, and an announcement generation method. [Background technology]

[0002] There are known devices that generate information to be announced to users. For example, Patent Document 1 discloses an announcement information generation device that includes an announcement information storage means that stores keywords and announcement information related to the keywords in association with each other. The announcement information generation device selects announcement information corresponding to the keywords from the announcement information storage means based on the keywords, adds the selected announcement information to generate a template, and generates an email based on the template. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2010-205044 Summary of the Invention [Problem to be solved by the invention]

[0004] Recently, a technology has been developed that generates sentences by inputting desired requirements into a trained model formed by learning a vast amount of linguistic information. However, even if requirements are input into such a trained model, the trained model may not output the information expected by the user. Currently, users use trained models through trial and error. As a result, users may waste time trying to obtain the desired information. [Means for solving the problem]

[0005] (1) A program for generating an announcement that solves the problems includes a first step of causing a computer to acquire announcement source information for creating the announcement, a second step of causing a language processing computer to generate prompt information for generating the announcement based on the announcement source information, and a third step of inputting the prompt information generated in the second step into the language processing computer and acquiring the announcement from the language processing computer. With this configuration, the announcement can be acquired with a small number of trial and error attempts.

[0006] (2) In the announcement generation program described in (1) above, in the second step, the computer is caused to generate the prompt information so as to include the announcement source information and common instruction information for causing the language processing capable computer to generate the announcement based on the announcement source information. With this configuration, the announcement can be obtained using the announcement source information.

[0007] (3) In the announcement generation program according to (1) or (2), in the second step, the computer is caused to generate the prompt information so as to include the announcement source information, system example information consisting of a set of prepared example announcement source information and an example announcement sentence associated with the example announcement source information, common instruction information for causing the language processing computer to generate the announcement sentence based on the announcement source information, and first instruction information for instructing the language processing computer to generate the announcement sentence in accordance with the system example information. With this configuration, an announcement sentence similar to the example announcement sentence can be obtained.

[0008] (4) In the announcement generation program described in (3) above, the system example information is information created based on information provided by a user. With this configuration, an announcement corresponding to the information provided by the user can be obtained.

[0009] (5) In the announcement generation program described in (3) above, the system example information is information created in advance for each business type or each user's affiliation. With this configuration, an announcement according to the business type or the user's affiliation can be obtained.

[0010] (6) In the announcement generation program according to any one of (1) to (5), in the second step, the prompt information is generated so as to further include second instruction information, the second instruction information being instruction information for instructing the language processing computer to create the announcement within a specified number of characters. With this configuration, an announcement of a desired number of characters can be obtained.

[0011] (7) In the announcement generation program according to any one of (1) to (6), in the second step, the prompt information is generated so as to include third instruction information, and the third instruction information is instruction information that instructs the language processing computer to prompt the user to input words that will contribute to the generation of the announcement if the announcement source information does not contain information that can be used to generate the announcement. This configuration can improve the completeness of the announcement.

[0012] (8) In the announcement generation program according to any one of (1) to (7), in the second step, the prompt information is generated so as to include fourth instruction information, and the fourth instruction information is instruction information for instructing the language processing computer to generate the announcement so as not to include any prohibited words. With this configuration, an announcement that does not include any prohibited words can be obtained.

[0013] (9) In the announcement generation program according to any one of (1) to (8), the announcement source information includes purpose information relating to the purpose of the announcement. With this configuration, an announcement according to the purpose of the announcement can be obtained.

[0014] (10) In the announcement generation program described in (9) above, the announcement source information further includes at least one of location information about the location where the announcement will be made, announcement time information about the time or time period when the announcement will be made, and event time information about the time or time period of an event included in the content of the announcement. With this configuration, an announcement including information about the location or time can be obtained.

[0015] (11) A system for generating announcements that solves the above problems includes a first processing unit that acquires announcement source information for creating the announcement, a second processing unit that generates prompt information for causing a language processing computer to generate the announcement based on the announcement source information, and a third processing unit that inputs the prompt information to the language processing computer and acquires the announcement from the language processing computer. With this configuration, the announcement can be acquired with a small number of trial and error attempts.

[0016] (12) A method for generating an announcement that solves the above problem is a computer-executed method for generating an announcement, the method including: a first step of acquiring announcement source information for creating an announcement; a second step of generating prompt information for causing a language processing computer to generate the announcement based on the announcement source information; and a third step of inputting the prompt information generated in the second step into the language processing computer and acquiring the announcement from the language processing computer. With this configuration, an announcement can be obtained with a small number of trial and error attempts. [Effects of the Invention]

[0017] According to the announcement generation program, announcement generation system, and announcement generation method, an announcement can be obtained with a small number of trial and error attempts. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram of an announcement generation system according to a first embodiment. [Figure 2] FIG. 2 is a diagram of an input screen of an input interface. [Figure 3] FIG. 10 is a block diagram of a first example of prompt information. [Figure 4] FIG. 10 is a block diagram of a second example of prompt information. [Figure 5] FIG. 10 is a diagram illustrating an example of instruction information. [Figure 6] FIG. 10 is a diagram of an output screen of the output interface. [Figure 7] FIG. 10 is a block diagram of an announcement generation system according to a second embodiment. [Figure 8] FIG. 10 is a block diagram of an announcement generation system according to a third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0019] First Embodiment An announcement generation system 1 according to the first embodiment will be described with reference to Figures 1 to 6. The announcement generation system 1 generates an announcement.

[0020] In this embodiment, an announcement is a sound source that is broadcast within a broadcast area. Examples of a broadcast area include an area set within a building, including a facility, an outdoor area outside a building on a premises, and a wide-ranging broadcast area such as a disaster prevention radio. An announcement includes background music, sound effects, and audio related to the broadcast content. Broadcast content includes announcements, weather information, greetings, etc. Announcements include guidance within the broadcast area, event information, product information, notifications about lost children, disasters, event times, and schedule delays. An announcement text is a text that corresponds to the broadcast content.

[0021] Conventionally, announcements are created by broadcasters using information found in information media. Information media include web content and books published on the Internet. Announcements are also created by editing previous announcements.

[0022] Incidentally, announcements are not just sentences, since they are information intended for the general public to hear. Announcements are created with various considerations in mind. For example, to ensure that announcements are understandable through audio, various considerations are taken into account, such as simple and clear language, word order, wording appropriate to the content of the announcement, a way of speaking that conforms to social conventions, wording appropriate to the location of the broadcast, and a way of speaking that is appropriate to the age group of people in the broadcast range. For this reason, creating appropriate announcements has been difficult or time-consuming. In this context, the announcement generation system 1 contributes to easily creating announcements.

[0023] [Announcement Generation System] The announcement generation system 1 generates an announcement.

[0024] 1, the announcement generation system 1 includes a computer 10. The computer 10 includes a first processing unit 11, a second processing unit 12, and a third processing unit 13. The announcement generation system 1 is communicatively connected to a language processing capable computer 2. The announcement generation system 1 is communicatively connected to an information terminal 3.

[0025] The first processing unit 11 acquires announcement source information A1 for creating an announcement. The second processing unit 12 generates prompt information P1 for causing the language processing computer 2 to generate an announcement based on the announcement source information A1. The third processing unit 13 inputs the prompt information P1 generated in the second step into the language processing computer 2 and acquires the announcement from the language processing computer 2.

[0026] The first processing unit 11, the second processing unit 12, and the third processing unit 13 are each constructed by a computational element of the computer 10 and an announcement sentence generation program installed in the computer 10. The first processing unit 11, the second processing unit 12, and the third processing unit 13 may be functions on a program realized within one computer 10. The first processing unit 11, the second processing unit 12, and the third processing unit 13 may also be configured by computers 10 that are independent of each other.

[0027] The computer 10 includes a personal computer, a portable computer, a server, or a general-purpose computer. Portable computers include notebook computers and smartphones. The computer 10 may be a control device incorporated into a broadcasting device or the like. The control device has a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) and a storage unit. The storage unit includes a semiconductor memory, a hard disk, an optical disk, and a magnetic medium.

[0028] [Announcement generation program] The announcement generation program (hereinafter simply referred to as the program) is an announcement generation program that generates an announcement. The program includes a first step, a second step, and a third step. The first step is a program element for realizing the first processing unit 11. The second step is a program element for realizing the second processing unit 12. The third step is a program element for realizing the third processing unit 13.

[0029] Steps are described as follows. Because a step is part of a program, it can be described as "a step is a step that causes the computer 10 to perform a predetermined process." In the following sentences, the phrase "is a step" will be omitted for the sake of brevity and clarity. Therefore, a step will be described as a process performed by a program. Specifically, it will be described as "In a step, the program causes the computer 10 to perform a predetermined process."

[0030] [Step 1] In a first step, the program causes the computer 10 to acquire announcement source information A1 for creating an announcement sentence. In one example, in the first step, the program causes the computer 10 to acquire the announcement source information A1 via the input interface 4.

[0031] The announcement source information A1 includes purpose information related to the purpose of the announcement. The announcement source information A1 further includes at least one of location information related to the location where the announcement will be made, announcement time information related to the time or time period when the announcement will be made, and event time information related to the time or time period of an event included in the content of the announcement. The announcement source information A1 may also include detailed content including detailed content of the announcement.

[0032] The input interface 4 is an information terminal 3 that can input information to the computer 10. Examples of the information terminal 3 are a smartphone, a personal computer, etc. The information terminal 3 has a display unit that can display an input screen 20 that can input the announcement source information A1.

[0033] 2 shows an example of the input screen 20. The input screen 20 is provided with a first input section 21, a second input section 22, and a third input section 23. The first input section 21 to the third input section 23 are configured as text boxes into which text can be input. The character information input into the first input section 21 to the third input section 23 is acquired by the computer 10 as announcement source information A1 in a first step.

[0034] The word "Location" is displayed around the first input unit 21 so that the location where the announcement will be made can be entered into the first input unit 21. The word "Purpose" is displayed around the second input unit 22 so that the purpose of the announcement can be entered into the second input unit 22. The word "Details" is displayed around the third input unit 23 so that detailed content of the announcement can be entered into the third input unit 23. Even if the information input into the first input unit 21 to the third input unit 23 is not distinguished, the language processing capable computer 2 can perform processing in accordance with instructions based on this information. For this reason, the program acquires the information input into the first input unit 21 to the third input unit 23 without distinguishing between them. Therefore, the first input unit 21 to the third input unit 23 may be configured as a single text box. Note that in this embodiment, the first input unit 21 to the third input unit 23 are not configured as a single text box but are divided into three. This is to prevent the user from missing any input. When generating an announcement, the location and purpose of the announcement are often required, and therefore a first input section 21 and a second input section 22 are provided. A third input section 23 is provided for inputting other detailed content. When inputting detailed information is required when generating an announcement, the input sections into which the announcement source information A1 can be input may be divided into four or more sections. Furthermore, the input sections into which the announcement source information A1 can be input may be provided across multiple screens.

[0035] [Second Step] The second step will now be described. In the second step, the program causes the computer 10 to generate prompt information P1 based on the announcement source information A1. The prompt information P1 is information for causing the language processing capable computer 2 to generate an announcement sentence.

[0036] The language processing capable computer 2 is a device that operates a large-scale language model that has been trained based on a huge amount of language information. The language processing capable computer 2 is connected to the computer 10 via a network. An example of a network is the Internet.

[0037] Examples of large-scale language models (also abbreviated as LLMs) include ChatGPT (registered trademark), Bard (registered trademark), LaMDA (registered trademark), LLaMA (registered trademark), tsuzumi (registered trademark), CyberAgentLM (registered trademark), etc. Large-scale language models do not have to be provided on a cloud, but may be constructed by the user themselves.

[0038] The large-scale language model generates an answer based on the prompt information P1. For example, if the prompt information P1 includes an instruction to create an announcement, the large-scale language model generates an announcement. The large-scale language model generates an answer according to the prompt information P1.

[0039] [First example of the second step] In the first example of this embodiment, in the second step, the program causes the computer 10 to generate prompt information P1 so that the prompt information P1 includes announcement source information A1 and common instruction information B1.

[0040] FIG. 3 shows a block diagram of a first example of prompt information P1. As shown in FIG. 3, prompt information P1 is configured as text information. Prompt information P1 is configured as a collection of announcement source information A1 and common instruction information B1. In prompt information P1, the common instruction information B1 is arranged first, followed by announcement source information A1. Note that the order in which each piece of information is written in prompt information P1 is not limited and can be determined arbitrarily.

[0041] [Common instruction information] The common instruction information B1 is information commonly used to make the language processing capable computer 2 generate announcements. The common instruction information B1 includes "You are an assistant who creates announcements for broadcasts," "Please generate announcements based on user input," and "Please behave in accordance with the following rules." Note that the notation of the common instruction information B1 is not limited to this.

[0042] [Second example of second step] In a second example of this embodiment, in a second step, the program causes the computer 10 to generate prompt information P1 so that the prompt information P1 includes announcement source information A1, common instruction information B1, system example information B2, and first instruction information C1. The announcement source information A1 is the information described above. The common instruction information B1 is the same as the information described in the first example.

[0043] 4 shows a block diagram of prompt information P1 of a second example. As shown in FIG. 4, prompt information P1 is configured as text information. Prompt information P1 is configured as a collection of announcement source information A1, system exemplification information B2, common instruction information B1, and first instruction information C1. In one example, in prompt information P1, the common instruction information B1, the first instruction information C1, the system exemplification information B2, and the announcement source information A1 are arranged in this order from front to back.

[0044] 5 shows examples of common instruction information B1, first instruction information C1, and system exemplification information B2. System exemplification information B2 is an example of an input to language processing computer 2 and an output of language processing computer 2 in response to that input, and is information for causing language processing computer 2 to operate in accordance with that example.

[0045] [System example information] The system example information B2 is information consisting of a set of example announcement source information B21 prepared in advance and an example announcement sentence B22 associated with the example announcement source information B21.

[0046] An example of the system example information B2 is as follows: In the example system information B2, the example announcement source information B21 includes "announcement to be broadcast in Store A," "notice of the Hokkaido product fair starting at 10:00 on February 2nd," and "the fair will be held in the underground event hall."

[0047] In the example system information B2, the example announcement sentence B22 includes, "Thank you very much for visiting Store A today. We would like to inform our customers about an event. On February 2nd, from 2 pm, we will be holding a Hokkaido products fair in the underground event space. We look forward to seeing you there."

[0048] The prompt information P1 may include a plurality of different pieces of system exemplification information B2. If the number of pieces of system exemplification information B2 is large, there is a risk that the output of the language processing capable computer 2 will diverge due to variations in the information about the system exemplification information B2. In this case, the user will not receive the desired response from the language processing capable computer 2.

[0049] The system example information B2 may be information created based on information provided by a user who uses the program (hereinafter referred to as user system example information B3). Retail businesses often have their own original announcements created in the past. For example, a shopping mall may have an announcement that reads, "Thank you for visiting Store A today. We would like to inform our customers of an upcoming event. We will be holding a Hokkaido Product Fair in the basement event space on February 2nd, starting at 2:00 PM. We look forward to seeing you there." In this case, the keywords "Store A," "Hokkaido," and "fair" extracted from the announcement are defined as example announcement source information B21. The announcement is defined as example announcement sentence B22. Then, information consisting of a pair of example announcement source information B21 and example announcement sentence B22 associated with example announcement source information B21 is stored in a storage unit as one of system example information B2.

[0050] The system example information B2 may be information created in advance for each industry (hereinafter referred to as industry-specific system example information B4). Examples of industries include retail, manufacturing, office work, education, transportation, finance, and medical care. Industries may be further subdivided. For example, retail may be subdivided into convenience stores, shopping malls, department stores, specialty stores, and the like. Education may be subdivided into elementary schools, junior high schools, universities, kindergartens, nursery schools, cram schools, driving schools, and the like.

[0051] The system example information B2 may be information created in advance for each organization to which the user belongs (hereinafter referred to as organization-specific system example information). Examples of organizations to which the user belongs include building management, tenants, individuals, and the like.

[0052] The system exemplification information B2 is prepared in advance. The system exemplification information B2 is stored in advance in the computer 10. The system exemplification information B2 may be stored in an external server. The system exemplification information B2 may be stored in the server in a manner separated for each user.

[0053] [First instruction information] The first instruction information C1 is information that instructs the language processing capable computer 2 to generate an announcement sentence in accordance with the system example information B2. An example of the first instruction information C1 is an instruction such as "The following describes the rules, including expected input and output formats." Note that the way in which the first instruction information C1 gives instructions is not limited to this.

[0054] By providing the first instruction information C1 together with the announcement source information A1 to the language processing capable computer 2, it is possible to obtain an output as a response from the language processing capable computer 2 that conforms to the input-output relationship between the input and output in the system exemplification information B2.

[0055] In this embodiment, in the section "The following describes the rules, including the expected input and output format," the "expected input" corresponds to the example announcement source information B21 in the system example information B2. The "output format" corresponds to the example announcement sentence B22 in the system example information B2.

[0056] [Second instruction information] In the second step, the program may further generate prompt information P1 so that second instruction information C2 is included. The second instruction information C2 is instruction information that instructs the language processing computer 2 to create an announcement within a specified number of characters. An example of the second instruction information C2 is an instruction such as "Please make the announcement to be output approximately 100 to 300 characters long." Note that the manner in which the second instruction information C2 is given is not limited to this.

[0057] [Third instruction information] In the second step, the program may further generate prompt information P1 so that third instruction information C3 is included. The third instruction information C3 is instruction information that instructs the language processing computer 2 to prompt the user to input words that will contribute to generating an announcement sentence if the announcement source information A1 does not contain information that can be used to generate an announcement sentence. An example of the third instruction information C3 is an instruction saying, "If the information necessary for the purpose is not included, please prompt." Note that the manner in which the third instruction information C3 is given is not limited to this.

[0058] [4th instruction information] In the second step, the program may further generate the prompt information P1 so that the prompt information P1 includes the fourth instruction information C4. The fourth instruction information C4 is instruction information that instructs the language processing computer 2 to generate an announcement sentence that does not include any prohibited words. An example of the fourth instruction information C4 is an instruction such as "Please create an announcement sentence that does not include the following prohibited words. Prohibited words: evil, death, suffering." Note that the way in which the fourth instruction information C4 is given is not limited to this.

[0059] [Third Step] In a third step, the program causes computer 10 to input the prompt information P1 formed in the second step to language processing computer 2. Then, the program causes computer 10 to obtain from language processing computer 2 an announcement sentence that language processing computer 2 outputs based on the input.

[0060] The program may cause the computer 10 to output the announcement sentence obtained from the language processing capable computer 2 to the output interface 5 in a third step. The output interface 5 is an information terminal 3 that can receive information from the computer 10. Examples of the information terminal 3 include a smartphone, a personal computer, etc. The information terminal 3 has a display unit that can display an output screen 30 on which announcements can be output. The output interface 5 may be the same device as the input interface 4.

[0061] 6 shows an output screen 30 that displays the announcement source information A1 and the announcement text generated by the language processing capable computer 2. This output screen 30 includes an input screen 20 for inputting the announcement source information A1. The input screen 20 and the output screen 30 may be configured as separate screens with different pages. The output screen 30 is provided with an output unit 31 that displays the announcement text.

[0062] When the output screen 30 includes the input screen 20, it is easy to compare the announcement source information A1 with the announcement text generated based on the announcement source information A1.

[0063] <Operation of the embodiment> The language processing capable computer 2 outputs various answers based on the prompt information P1. If the prompt information P1 is inappropriate, the desired answer cannot be obtained. In this regard, the program according to this embodiment causes the computer 10 to generate the prompt information P1 based on the announcement source information A1. Therefore, even a person who has never used the language processing capable computer 2 or a person who does not understand the commands to give to the language processing capable computer 2 can obtain an announcement by having the announcement generation system 1 or the program obtain the announcement source information A1. In one example, by having the announcement generation system 1 or the program obtain keywords related to the announcement, an announcement containing the keywords can be obtained.

[0064] <Effects of the embodiment> (1) The program includes first to third steps. In the first step, the computer 10 acquires announcement source information A1 for creating an announcement. In the second step, the computer 10 generates prompt information P1 for causing the language processing capable computer 2 to generate an announcement, based on the announcement source information A1. In the third step, the computer 10 inputs the prompt information P1 generated in the second step into the language processing capable computer 2, and acquires the announcement from the language processing capable computer 2. With this configuration, the announcement can be acquired with a small number of trial and error iterations.

[0065] (2) In a second step, the program causes the computer 10 to generate prompt information P1 so that the prompt information P1 includes the announcement source information A1 and common instruction information B1. The common instruction information B1 is instruction information for causing the language processing computer 2 to generate an announcement sentence based on the announcement source information A1. With this configuration, the announcement sentence can be obtained from the announcement source information A1.

[0066] (3) In a second step, the program causes the computer 10 to generate prompt information P1 so that the prompt information P1 includes the announcement source information A1, the system example information B2, the common instruction information B1, and the first instruction information C1. With this configuration, an announcement sentence similar to the example announcement sentence B22 can be obtained.

[0067] (4) The program includes system example information B2, which is information created based on information provided by a user who uses the program. With this configuration, an announcement text corresponding to the information provided by the user can be obtained.

[0068] (5) In the program, the system example information B2 is information created in advance for each business type or each user's affiliation. With this configuration, an announcement text according to the business type or the user's affiliation can be acquired.

[0069] (6) In a second step, the program further generates prompt information P1 so that second instruction information C2 is included. The second instruction information C2 is instruction information that instructs the language processing computer 2 to create an announcement within a specified number of characters. With this configuration, an announcement of the desired number of characters can be obtained.

[0070] (7) In the second step, the program further generates prompt information P1 so that third instruction information C3 is included. The third instruction information C3 is instruction information that instructs the language processing computer 2 to prompt the user to input words that will contribute to the generation of an announcement if the announcement source information A1 does not contain information that can be used to generate an announcement. This configuration can improve the completeness of the announcement.

[0071] (8) In the second step, the program may further generate prompt information P1 so that fourth instruction information C4 is included. The fourth instruction information C4 is instruction information that instructs the language processing computer 2 to generate an announcement that does not include prohibited words. This configuration makes it possible to obtain an announcement that does not include prohibited words. This allows the announcement to be formed into a more appropriate sentence.

[0072] (9) The announcement source information A1 includes purpose information about the purpose of the announcement. With this configuration, an announcement text according to the purpose of the announcement can be obtained.

[0073] (10) The announcement source information A1 further includes at least one of location information about the location where the announcement will be made, announcement time information about the time or time period when the announcement will be made, and event time information about the time or time period of an event included in the content of the announcement. With this configuration, it is possible to obtain an announcement text that includes information about the location or time.

[0074] (11) The announcement generation system 1 includes a first processing unit 11 that acquires announcement source information A1 for creating an announcement, a second processing unit 12, and a third processing unit 13. The second processing unit 12 generates prompt information P1 for causing the language processing capable computer 2 to generate an announcement based on the announcement source information A1. The third processing unit 13 inputs the prompt information P1 to the language processing capable computer 2 and acquires the announcement from the language processing capable computer 2. With this configuration, an announcement can be acquired with a small number of trial and error attempts.

[0075] (12) The announcement generation method is a method for generating an announcement executed by a computer 10. The announcement generation method includes a first step, a second step, and a third step. In the first step, the computer 10 acquires announcement source information A1 for creating an announcement. In the second step, the computer 10 generates prompt information P1 for causing the language processing capable computer 2 to generate an announcement based on the announcement source information A1. In the third step, the computer 10 inputs the prompt information P1 generated in the second step into the language processing capable computer 2 and acquires the announcement from the language processing capable computer 2. With this configuration, an announcement can be acquired with a small number of trial and error attempts.

[0076] Second Embodiment An announcement generation system 1 according to the second embodiment will be described with reference to Fig. 7. In this embodiment, components common to the first embodiment are assigned the same reference numerals as those in the first embodiment, and descriptions of overlapping components will be omitted.

[0077] In the announcement generation system 1 of the first embodiment, information prepared in advance is used as the system example information B2. In contrast, in the present embodiment, information provided by the user is used as the system example information B2. The following describes changes made to the announcement generation system 1 from the first embodiment.

[0078] The announcement generation system 1 is communicatively connected to a language processing capable computer 2. The announcement generation system 1 is communicatively connected to an information terminal 3.

[0079] Furthermore, the announcement generation system 1 is connected to a first server 7 directly or via a network. The first server 7 may be provided within the announcement generation system 1.

[0080] The first server 7 stores user system example information B3 as system example information B2. The user system example information B3 is a type of system example information B2 created based on information provided by a user using the program. The user system example information B3 may be information created by the user himself / herself. The user system example information B3 may also be information created by the user through use of the program of this embodiment. The user system example information B3 is stored in the first server 7 in a state associated with user account information.

[0081] The first processing unit 11 executes the same processing as that shown in the first embodiment. The second processing unit 12 executes the same processing as that shown in the first embodiment. Specifically, the second processing unit 12 forms prompt information P1 for causing the language processing capable computer 2 to generate an announcement sentence based on the announcement source information A1. Then, the second processing unit 12 generates the prompt information P1 so that the user system example information B3 is included.

[0082] [Announcement generation program] The announcement generation program (hereinafter simply referred to as the program) is an announcement generation program that generates an announcement. The program causes the computer 10 to obtain user account information of a user who uses the program. In one example, the program causes the computer 10 to obtain user account information of a user who uses the program when the program is started.

[0083] In a first step, the program causes the computer 10 to acquire announcement source information A1 for creating an announcement. The user account information includes user information and information about the industry to which the user belongs. In the first step, the program causes the computer 10 to acquire user system example information B3 associated with the user information based on the user account information.

[0084] In the second step, the program causes the computer 10 to generate prompt information P1 so that the prompt information P1 includes the announcement source information A1, the user system example information B3 as the system example information B2, the common instruction information B1, and the first instruction information C1. As a result, the user system example information B3 is automatically included in the prompt information P1.

[0085] In a third step, the program causes the prompt information P1 formed in the second step to be input to the language processing computer 2. Then, the program causes the computer 10 to obtain from the language processing computer 2 an announcement sentence that the language processing computer 2 outputs based on the input.

[0086] <Effects of the embodiment> In the second step, the program generates prompt information P1 so that the prompt information P1 includes user system example information B3. With this configuration, an announcement is generated based on the announcement source information A1 and the user system example information B3. This allows the user system example information B3 to be reflected in the newly generated announcement.

[0087] <Third embodiment> An announcement generation system 1 according to the third embodiment will be described with reference to Fig. 8. In this embodiment, components common to the first embodiment are assigned the same reference numerals as those in the first embodiment, and descriptions of overlapping components will be omitted.

[0088] In the announcement generation system 1 of the first embodiment, information prepared in advance is used as the system example information B2. In contrast, in the present embodiment, industry-specific system example information B4 is used as the system example information B2. The following describes changes made to the announcement generation system 1 from the first embodiment.

[0089] The announcement generation system 1 is communicatively connected to a language processing capable computer 2. The announcement generation system 1 is communicatively connected to an information terminal 3.

[0090] Furthermore, the announcement generation system 1 is connected to a second server 8 directly or via a network. The second server 8 may be configured integrally with the first server 7 of the second embodiment. The first server 7 and the second server 8 may also be provided within the announcement generation system 1.

[0091] The second server 8 stores industry-specific system example information B4 as system example information B2. The industry-specific system example information B4 is managed by industry. The industry-specific system example information B4 is configured to be searchable by a program based on industry information. The industry-specific system example information B4 may be information created based on information provided by a user using the program. The industry-specific system example information B4 may be information created by the user. The industry-specific system example information B4 may also be information created by the user using the program of this embodiment. The industry-specific system example information B4 is stored in the second server 8 in a state associated with industry parameters that distinguish industries.

[0092] The first processing unit 11 executes the same processing as that shown in the first embodiment. The second processing unit 12 executes the same processing as that shown in the first embodiment. Specifically, the second processing unit 12 forms prompt information P1 for causing the language processing capable computer 2 to generate an announcement sentence based on the announcement source information A1. Then, the second processing unit 12 generates the prompt information P1 so that the industry-specific system example information B4 is included.

[0093] [Announcement generation program] The announcement generation program (hereinafter simply referred to as the program) is an announcement generation program that generates an announcement. The program causes the computer 10 to obtain user account information of a user who uses the program. In one example, the program causes the computer 10 to obtain user account information of a user who uses the program when the program is started.

[0094] In a first step, the program causes the computer 10 to acquire announcement source information A1 for creating an announcement. The user account information includes user information and information on the industry to which the user belongs. In the first step, the program causes the computer 10 to acquire industry-specific system example information B4 associated with the user's industry information based on the user account information.

[0095] In the second step, the program causes the computer 10 to generate prompt information P1 so as to include announcement source information A1, industry-specific system example information B4 as system example information B2, common instruction information B1, and first instruction information C1. As a result, the industry-specific system example information B4 is automatically included in the prompt information P1.

[0096] In a third step, the program causes the prompt information P1 formed in the second step to be input to the language processing computer 2. Then, the program causes the computer 10 to obtain from the language processing computer 2 an announcement sentence that the language processing computer 2 outputs based on the input.

[0097] <Effects of the embodiment> In the second step, the program generates prompt information P1 so that the prompt information P1 includes the industry-specific system example information B4. With this configuration, an announcement is generated based on the announcement source information A1 and the industry-specific system example information B4. This allows the industry-specific system example information B4 to be reflected in the newly generated announcement.

[0098] <Modification> The above embodiments are examples of forms that the announcement generation program, announcement generation system 1, and announcement generation method can take, and are not intended to limit the forms. The announcement generation program, announcement generation system 1, and announcement generation method can take forms different from those exemplified in the above embodiments. Examples include forms in which part of the configuration of the embodiments is replaced, changed, or omitted, or forms in which new configuration is added to the embodiments. Modified examples of the embodiments are shown below.

[0099] In the announcement generation system 1 of the second embodiment, the system exemplification information B2 used to generate the prompt information P1 may be configured to be selectable by a user operation. In one example, the system exemplification information B2, which is preset as in the first embodiment, and the user system exemplification information B3 are configured to be selectable by a user operation.

[0100] In the announcement generation system 1 of the third embodiment, the system example information B2 used to generate the prompt information P1 may be configured to be selectable by a user operation. In one example, the system example information B2 preset as in the first embodiment and the industry-specific system example information B4 are configured to be selectable by a user operation.

[0101] In the third embodiment, the program sets the business type information based on the user account information. Alternatively, the program may set the business type information based on the user's selection. In this configuration, the user can use information on other businesses when creating an announcement without being limited to the user's own business type registered as user account information.

[0102] In the third embodiment, the industry-specific system example information B4 may be information including the affiliation and the industry. Examples of industries include retail and manufacturing. Examples of affiliation include building management, tenants, and individuals. Examples of information including the affiliation and the industry include retail building management, retail tenants, retail individuals, manufacturing building management, manufacturing tenants, and manufacturing individuals. In this case, the system example information B2 is created in advance according to each piece of information including the affiliation and the industry. By subdividing the system example information B2, it is possible to obtain announcement text according to the user's industry and affiliation.

[0103] In the second embodiment, the program may cause the computer 10 to learn the editing characteristics of the user based on the user log information. The program may cause the computer 10 to edit the announcement generated by the language processing capable computer 2 based on the learning results, and then output the edited announcement. The learning content is the selection of words, expressions, and polite language. Examples of learning results are converting "minna-san" (everyone) to "your customer," converting "shimasu" (will do) to "itashimasu," converting "yessssss" (please) to "yessssssssss," converting "nobimasu" (will go up) to "nobimasu" (will go up), etc.

[0104] In the first to third embodiments, the language processing capable computer 2 is a device that operates a large-scale language model, but the language processing capable computer 2 may be any device that can process language. For example, the language processing capable computer 2 may have a language processing model that operates on a rule basis. In one example of a rule base, an announcement is created by selecting a template for the announcement and inserting keywords into the selected template.

[0105] The program may cause the computer 10 to generate an announcement based on a rule. The announcement generated based on the rule may be used as the system example information B2 in the first to third embodiments. In this case, the announcement is provisionally generated based on the rule and then refined using a large-scale language model. In this way, the announcement is generated using a hybrid of the rule base and the large-scale language model, thereby suppressing the divergence of the generated announcement.

[0106] Although the first to third embodiments have been described using Japanese as an example, the technology disclosed herein can also be applied to generating announcements in foreign languages. The technology disclosed herein is useful when generating announcements in multiple languages. Specifically, the announcement generation system 1 generates announcements in the languages ​​of specified countries by acquiring announcement source information A1 in Japanese. The announcement generation system 1, which generates announcements in multiple languages, prepares system example information B2, industry-specific system example information B4, or user system example information B3 for each country. The system example information B2 and industry-specific system example information B4 include example announcements B22 that conform to the customs of each country. This allows the generated announcements to conform to the customs of each country. Therefore, more appropriate announcements can be generated than when Japanese announcements are simply machine-translated.

[0107] In the second embodiment, the number of input units may be increased or decreased based on user account information. Since the number of keywords used in announcements varies depending on the user, such a configuration makes it possible to generate announcements suited to the user. In this case, the number of input units into which the announcement source information A1 is input is linked in advance to the user account information. The number of input units may be configured to be changeable by user operation.

[0108] In the second embodiment, the program may cause the computer 10 to analyze the relationship between the user account information and the announcement source information A1. Furthermore, the program may cause the computer 10 to form characteristics of how the announcement source information A1 is input (hereinafter referred to as input characteristics) for the user associated with the user account information based on the analysis results. In short, the program may cause the computer 10 to find out the user's habits from the announcement source information A1. Then, the program may cause the computer 10 to output advice to the user based on the user's input characteristics when the user makes an input. Examples of input characteristics are failure to input a location or failure to input a date and time. If the user has such input characteristics, the computer 10 outputs messages such as "Please input a date and time" or "Please input a location" to the input interface 4.

[0109] In the first to third embodiments, the program may cause the computer 10, when the announcement source information A1 contains a keyword, to check whether any related keywords related to the keyword have been omitted from the announcement source information A1. Furthermore, the program may cause the computer 10 to output a prompt message to the input interface 4 to prompt the user to enter the related keywords if any related keywords have been omitted from the announcement source information A1. An example of a "keyword" is "lost child." Related keywords include "age," "years old," "male," "female," "elderly," "infant," and "child." The announcement generation system 1 outputs a prompt message when the announcement source information A1 contains a keyword and the announcement source information A1 is missing a related keyword. This makes it easier to generate a nearly complete announcement with no missing information in the announcement content when the announcement source information A1 contains a lost child.

[0110] In the first to third embodiments, the program may cause the computer 10 to add comments to words in the announcement. The comments are added to the announcement so that they are displayed on the screen when a predetermined word in the announcement is hovered over with the mouse while the announcement is displayed on the output interface 5. For example, if the announcement contains the word "lost," a comment such as "Have you specified the gender?" is added to "lost." This allows the user to know that information has been omitted from the announcement. [Explanation of symbols]

[0111] A1...announcement source information, B1...common instruction information, B2...system example information, B21...example announcement source information, B22...example announcement sentence, B3...user system example information, C1...first instruction information, C2...second instruction information, C3...third instruction information, C4...fourth instruction information, P1...prompt information, 1...announcement sentence generation system, 2...language processing capable computer, 4...input interface, 10...computer, 11...first processing unit, 12...second processing unit, 13...third processing unit.

Claims

1. An announcement generation program for generating an announcement, On the computer, a first step of acquiring announcement source information for creating the announcement; a second step of forming prompt information for causing a language processing capable computer to generate the announcement sentence based on the announcement source information; a third step of inputting the prompt information formed in the second step into the language processing computer and acquiring the announcement sentence from the language processing computer; Announcement text generator.

2. In the second step, the computer generating the prompt information so as to include the announcement source information and common instruction information for causing the language processing capable computer to generate the announcement sentence based on the announcement source information; 2. The announcement generation program according to claim 1.

3. In the second step, the computer The announcement source information; system example information consisting of a set of prepared example announcement source information and an example announcement sentence associated with the example announcement source information; common instruction information for causing the language processing capable computer to generate the announcement sentence based on the announcement source information; generating the prompt information to include first instruction information instructing the language-processing-enabled computer to generate the announcement sentence in accordance with the system instantiation information; 2. The announcement generation program according to claim 1.

4. The system example information is information created based on information provided by a user.

4. The announcement generation program according to claim 3.

5. The system example information is information created in advance for each business type or each user's affiliation.

4. The announcement generation program according to claim 3.

6. In the second step, the prompt information is generated so as to further include second instruction information; the second instruction information is instruction information instructing the language processing computer to create the announcement sentence within a specified number of characters; 6. The announcement generation program according to claim 1.

7. In the second step, the prompt information is generated so as to further include third instruction information; the third instruction information is instruction information for instructing the language processing computer to prompt a user to input a word that will contribute to generating the announcement sentence when the announcement source information does not contain information that can be used to generate the announcement sentence.

6. The announcement generation program according to claim 1.

8. In the second step, the prompt information is generated so as to further include fourth instruction information; the fourth instruction information is instruction information instructing the language processing capable computer to generate the announcement sentence so as not to include a prohibited word; 6. The announcement generation program according to claim 1.

9. The announcement source information includes purpose information regarding the purpose of the announcement.

6. The announcement generation program according to claim 1.

10. The announcement source information further includes: location information regarding where the announcement will be made; announcement time information regarding the time or time period when the announcement will be made; and event time information relating to the time or time period of the event included in the content of the announcement; 10. The announcement generation program according to claim 9.

11. An announcement generation system that generates an announcement, a first processing unit that acquires announcement source information for creating the announcement; a second processing unit that generates prompt information for causing a language processing computer to generate the announcement sentence based on the announcement source information; a third processing unit that inputs the prompt information to the language processing capable computer and acquires the announcement sentence from the language processing capable computer; Announcement text generation system.

12. A computer-implemented method for generating announcements, comprising: A first step of acquiring announcement source information for creating an announcement text; a second step of forming prompt information for causing a language processing capable computer to generate the announcement sentence based on the announcement source information; a third step of inputting the prompt information formed in the second step into the language processing computer and acquiring the announcement sentence from the language processing computer; Announcement text generation method.

Citation Information

Patent Citations

  • Apparatus, method and program for generating announcement information

    JP2010205044A