Lyrics generation device, lyrics generation method, and program

Abstract

We provide new technologies for delivering news. [Solution] The lyrics generation device according to this disclosure includes: acquisition means for acquiring news data representing the content of news; first prompt generation means for generating a first prompt using the news data that instructs the generation of song lyrics based on a summary of the news, keywords of the news, or both; and lyrics generation means for generating lyrics data representing lyrics by inputting the first prompt into a first generation model.

Description

Technical Field

[0001] The present disclosure relates to a lyric generation device, a lyric generation method, and a program.

Background Art

[0002] Technologies for providing news have been developed. For example, Patent Document 1 discloses a technique for outputting vocal sounds singing news and singing sounds with accompaniment by performing singing synthesis using news text as lyrics.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the invention of Patent Document 1, only singing sounds that sing the news content as it is can be generated. The present disclosure has been made in view of this problem, and one of its purposes is to provide a new technology for providing news.

Means for Solving the Problems

[0005] The lyric generation device according to the present disclosure includes an acquisition unit that acquires news data representing the content of news, a first prompt generation unit that generates a first prompt for instructing to generate lyrics of a music piece based on the summary of the news, the keyword of the news, or both of these, using the news data, and a lyric generation unit that generates lyric data representing the lyrics by inputting the first prompt into a first generation model.

[0006] The lyrics generation method relating to this disclosure is performed by a computer. The lyrics generation method includes an acquisition step of acquiring news data representing the content of a news article; a first prompt generation step of generating a first prompt using the news data that instructs the generation of song lyrics based on a summary of the news article, keywords of the news article, or both; and a lyrics generation step of generating lyrics data representing the lyrics by inputting the first prompt into a first generation model.

[0007] The program relating to this disclosure causes a computer to execute an acquisition step of acquiring news data representing the content of a news article; a first prompt generation step of generating a first prompt using the news data that instructs the computer to generate song lyrics based on a summary of the news article, keywords of the news article, or both; and a lyrics generation step of generating song lyrics data representing the song lyrics by inputting the first prompt into a first generation model. [Effects of the Invention]

[0008] According to this disclosure, new technologies for delivering news will be provided. [Brief explanation of the drawing]

[0009] [Figure 1] This diagram illustrates the general operation of the lyric generation device. [Figure 2] This is a block diagram illustrating the functional configuration of a lyric generation device. [Figure 3] This is a block diagram illustrating the hardware configuration of a computer that implements a lyric generation device. [Figure 4] This is a flowchart illustrating the processing flow performed by the lyric generation device. [Figure 5] This diagram illustrates a template for the first prompt. [Figure 6] This is a diagram illustrating schedule information. [Figure 7]This diagram illustrates the general operation of the lyric generation device. [Figure 8] This is a block diagram illustrating the functional configuration of a lyric generation device. [Figure 9] This is a flowchart illustrating the processing flow performed by the lyric generation device. [Modes for carrying out the invention]

[0010] Embodiments of the present disclosure will be described in detail below with reference to the drawings. In each drawing, the same or corresponding elements are denoted by the same reference numerals, and redundant explanations are omitted as necessary for clarity. Unless otherwise specified, predetermined values ​​such as specified values ​​and thresholds are stored in advance in a storage device accessible from the device that uses those values. Furthermore, unless otherwise specified, the storage unit is composed of one or any number of storage devices.

[0011] [Embodiment 1] <Overview> Figure 1 is a diagram illustrating an overview of the operation of the lyric generation device 2000. Herein, Figure 1 is intended to facilitate understanding of the overview of the lyric generation device 2000, and the operation of the lyric generation device 2000 is not limited to what is shown in Figure 1.

[0012] The lyrics generation device 2000 uses news data 10 to generate lyrics data 30 that represent lyrics based on the content of the news. News data 10 is text data that represents the content of the news. Lyrics data 30 includes text data of lyrics based on the content of the news.

[0013] The lyrics generation device 2000 generates lyrics data 30 using a first generation model 40. The first generation model 40 is configured to output text representing lyrics in response to a prompt instructing the generation of lyrics. The first generation model 40 may be a generation model specialized for generating lyrics, or it may be a general-purpose generation model capable of generating various types of data in response to requests.

[0014] The lyric generation device 2000 generates the first prompt 20 using the news data 10. The first prompt 20 is a prompt for instructing the generation of lyrics using the summary of the news shown in the news data 10, the keyword of the news, or both. Then, the lyric generation device 2000 generates lyric data 30 by inputting the first prompt 20 into the first generation model 40.

[0015] <Example of effects> According to the lyric generation device 2000 of the present embodiment, lyric data 30 is generated based on the summary of the news shown in the news data 10, the keyword of the news, or both. By generating the lyric data 30 based on the summary of the news, the keyword of the news, or both, the lyric data 30 can be generated by focusing on the important parts of the news. Therefore, for example, the lyric data 30 can be generated without being overly dependent on the specific content of the news shown in the news data 10.

[0016] Hereinafter, the lyric generation device 2000 of the present embodiment will be described in more detail.

[0017] <Example of functional configuration> FIG. 2 is a block diagram illustrating the functional configuration of the lyric generation device 2000. The lyric generation device 2000 includes an acquisition unit 2020, a first prompt generation unit 2040, and a lyric generation unit 2060. The acquisition unit 2020 acquires the news data 10. The first prompt generation unit 2040 generates the first prompt 20 using the news data 10. The lyric generation unit 2060 generates the lyric data 30 by inputting the first prompt 20 into the first generation model 40.

[0018] <Example of hardware configuration> Each functional component of the lyric generation device 2000 is implemented, for example, by hardware that realizes each functional component. Here, the hardware that realizes each functional component is, for example, a hardwired electronic circuit. In addition, each functional component of the lyric generation device 2000 is implemented by a combination of hardware and software. Here, a specific example of a combination of hardware and software is a combination of an electronic circuit and a program that controls it.

[0019] Figure 3 is a block diagram illustrating the hardware configuration of computer 1000, which implements the lyric generation device 2000. Computer 1000 is any computer. For example, computer 1000 is a stationary computer such as a PC (Personal Computer) or a server machine. Alternatively, computer 1000 is a portable computer such as a smartphone or a tablet device. Alternatively, computer 1000 is an integrated circuit such as a SoC (System on Chip). Computer 1000 may be a dedicated computer designed to implement the lyric generation device 2000, or it may be a general-purpose computer.

[0020] For example, by installing a predetermined application on computer 1000, the various functions of the lyric generation device 2000 are realized on computer 1000. The above application consists of a program for realizing each functional component of the lyric generation device 2000.

[0021] The method of obtaining the above program is arbitrary. For example, the program can be obtained from the storage medium on which it is stored. The storage medium on which the program is stored can be any storage medium, such as a DVD (Digital Versatile Disk) or a USB (Universal Serial Bus) memory. Alternatively, the program can be obtained by downloading it from a server device that manages the storage device on which the program is stored.

[0022] Computer 1000 includes a bus 1020, a processor 1040, memory 1060, a storage device 1080, an input / output interface 1100, and a network interface 1120. The bus 1020 is a data transmission path for the processor 1040, memory 1060, storage device 1080, input / output interface 1100, and network interface 1120 to send and receive data from each other. However, the method of connecting the processor 1040 and other components is not limited to bus connection.

[0023] The processor 1040 is a variety of computing devices such as a CPU (Central Processing Unit), MPU (Microprocessor Unit), GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), or DSP (Digital Signal Processor). The memory 1060 is the main memory, implemented using RAM (Random Access Memory), etc. The storage device 1080 is the auxiliary storage, implemented using a hard disk, SSD (Solid State Drive), memory card, or ROM (Read Only Memory), etc.

[0024] The input / output interface 1100 is an interface for connecting the computer 1000 with input / output devices. For example, input devices such as keyboards and output devices such as display devices are connected to the input / output interface 1100.

[0025] The network interface 1120 is an interface for connecting computer 1000 to a network. This network may be a LAN (Local Area Network) or a WAN (Wide Area Network).

[0026] The storage device 1080 stores programs that implement each functional component of the lyric generation device 2000 (programs that implement the aforementioned applications). The processor 1040 reads these programs into memory 1060 and executes them to implement each functional component of the lyric generation device 2000.

[0027] The lyric generation device 2000 may be implemented using multiple computers 1000. In this case, these multiple computers 1000 are interconnected by a network, bus, or the like. The configurations of these multiple computers 1000 may be identical or different.

[0028] <Processing flow> Figure 4 is a flowchart illustrating the processing flow performed by the lyric generation device 2000. The acquisition unit 2020 acquires news data 10 (S102). The first prompt generation unit 2040 generates a first prompt 20 using the news data 10 (S104). The lyric generation unit 2060 generates lyric data 30 by inputting the first prompt 20 to the first generation model 40 (S106).

[0029] <News data acquisition 10: 102> The acquisition unit 2020 acquires the news data 10 (S102). There are various ways to acquire the news data 10. For example, the news data 10 is transmitted from another device to the lyric generation device 2000. In this case, the acquisition unit 2020 acquires the news data 10 by receiving the news data 10 transmitted from the other device.

[0030] The device that transmits news data 10 to the lyrics generation device 2000 is, for example, a terminal used by a user of the lyrics generation device 2000 (hereinafter referred to as a user terminal). The user terminal is any computer, such as a smartphone or a PC.

[0031] For example, the lyrics generation device 2000 provides an input screen for entering news to the user terminal. The input screen is provided, for example, as a web page. The user enters text representing the news into the input screen displayed on the user terminal. The user terminal sends the text entered into the input screen to the lyrics generation device 2000. The acquisition unit 2020 acquires the text sent from the user terminal as news data 10.

[0032] The acquisition unit 2020 may accept input information that identifies a news item and acquire that news item. Information that identifies a news item may include, for example, the URL (Uniform Resource Locator) of the web page where the news item is published.

[0033] For example, the acquisition unit 2020 provides the user terminal with an input screen in which information identifying the news item can be entered. The acquisition unit 2020 acquires the text of the news item identified by the information entered on the input screen as news data 10.

[0034] The acquisition unit 2020 may automatically acquire the news data 10. For example, the acquisition unit 2020 acquires the news data 10 by searching for news information. For example, searching for news information can be achieved by accessing a predetermined web page where news information is provided and obtaining the news information. Alternatively, for example, searching for news information can be achieved by crawling the internet and collecting news information. Alternatively, for example, searching for news information can be achieved by accessing a database where news information is recorded and obtaining the news information.

[0035] There are various ways to obtain news data 10 by searching for news information. For example, the acquisition unit 2020 searches for news information at a predetermined date and time each day and obtains news data 10. As a more specific example, the acquisition unit 2020 obtains the information of the news ranked number 1 in the news ranking at the time the search is performed as news data 10. Alternatively, the acquisition unit 2020 may obtain the information of each news item that is ranked above a predetermined rank in the ranking as news data 10. When multiple news data 10 are obtained in this way, the lyrics generation device 2000 generates lyrics data 30 for each of these multiple news data 10.

[0036] The frequency at which news information is searched is at the user's discretion and is not limited to once a day.

[0037] The news shown in news data 10 may be limited to news of a predetermined genre. For example, suppose the predetermined genre is "computer security." In this case, news data 10 will be obtained for news related to computer security. For example, if news data 10 is obtained by searching for news information, the acquisition unit 2020 will search only for news of the predetermined genre. As mentioned above, if ranking information is used, news data 10 will be obtained using rankings for news of the predetermined genre.

[0038] The genres of news data 10 may be predetermined or can be specified by the user. For example, if a user uses a news application, the user's preferred genres may be identified based on the user's usage history of that application. In this case, the acquisition unit 2020 obtains news data 10 for news in the user's preferred genres.

[0039] <Generating the first prompt 20: S104> The first prompt generation unit 2040 generates a first prompt 20 using the news data 10 (S104). The first prompt 20 is a prompt that instructs the generation of lyrics using a summary of the news, keywords of the news, or both, for the news shown in the news data 10.

[0040] For example, the first prompt 20 consists of a first instruction text representing a phrase that instructs the generation of lyrics, and a second instruction text representing various information to be used in generating the lyrics. The first instruction text is text that represents instructions for generating lyrics, such as, "Please write lyrics using the information shown below."

[0041] The second instruction text includes at least the text of a summary or keywords of the news represented by the news data 10. If the second instruction text includes a summary of the news data 10, the first prompt generation unit 2040 generates a summary of the news data 10 using the news data 10. The first prompt generation unit 2040 then includes the generated summary of the news data 10 in the first prompt 20.

[0042] Here, various processes can be employed in the process of generating a text summary. If the first generation model 40 is a generation model capable of generating a text summary, the first prompt generation unit 2040 may generate a summary of the news data 10 by inputting the news data 10 into the first generation model 40.

[0043] If the second instruction text includes news keywords, the first prompt generation unit 2040 extracts one or more keywords from the news data 10. Then, the first prompt generation unit 2040 includes each of the extracted keywords in the first prompt 20.

[0044] Here, various processes can be employed for extracting keywords from text. If the first generation model 40 is a generation model that can extract keywords from text, the first prompt generation unit 2040 may extract keywords from the news data 10 by inputting the news data 10 into the first generation model 40.

[0045] The second instruction text may include information other than the summary and keywords of the news data 10. For example, the second instruction text may include a specification of the intention for writing the lyrics. By providing the intention for writing the lyrics to the first generative model 40, the first generative model 40 can be made to write lyrics that take the intention into consideration. The intention for writing the lyrics may be, for example, "to convey the content of the news concisely," "to convey the trends that can be obtained from the news," or "to convey the lessons that can be obtained from the news."

[0046] In addition, the second instruction text may include specifications for words or frames that should not be included in the lyrics. Hereafter, words and phrases will be collectively referred to as "words, etc." Also, words, etc. that should not be included in the lyrics will be referred to as "inappropriate words." By including inappropriate words in the second instruction text, it is possible to prevent inappropriate words and phrases from being included in the lyrics data 30.

[0047] Inappropriate words may be specified by specific words or by type. In the latter case, for example, inappropriate words may be proper nouns, company names, corporate names, or personal names. In addition, inappropriate words may also be specified by the function of words, such as "words or phrases that can identify an individual."

[0048] In addition, for example, the second instruction text includes specifications for various parameters related to the music. Hereafter, these parameters related to the music will also be called music parameters.

[0049] A song parameter is, for example, the section structure of a song. A song consists of one or more sections, such as an intro, verse, chorus, or outro. Therefore, the section structure of a song can be specified by listing the sections in the second instruction text, for example, as shown below. Intro Verse 1 Verse 2 · Chorus Outro

[0050] Other parameters include, for example, the genre of the song. Genres can be J-pop, K-pop, rock, Hawaiian, or Vocaloid.

[0051] Other examples of song parameters include the tempo of the song. The tempo of a song can be expressed quantitatively or qualitatively. Quantitative expressions of song tempo include, for example, the BPM (Beats per Minute) value or the BPM range. Qualitative expressions of song tempo include, for example, "very fast," "fast," "normal," "slow," or "very slow."

[0052] Other parameters include, for example, the mood of a song. The mood of a song represents the atmosphere of the song. The mood of a song can be expressed by a type of mood, such as upbeat, emotional, or energetic.

[0053] Other examples include song parameters, which represent the song's scene. A song's scene represents a scene suitable for that song. A song's scene can be represented by a type of scene, such as a club or a seaside.

[0054] The second instruction text may include arbitrary constraints in addition to the various constraints mentioned above. For example, the second instruction text may include a constraint such as, "Consider the news summary as context." Including such a constraint allows for lyrics to be created by interpreting the news more abstractly, without relying too heavily on information specific to the news.

[0055] The first prompt 20 is generated, for example, using a template. The template is pre-stored in a memory unit accessible from the lyrics generation device 2000. Here, multiple types of templates may be provided. In this case, the first prompt generation unit 2040 may accept a template specification from the user.

[0056] Figure 5 illustrates a template for the first prompt 20. The template 100 shown in Figure 5 includes a first instruction text 80 and a second instruction text 90. In Figure 5, the first instruction text 80 is set to fixed text. The second instruction text 90 includes, for each of several items, text representing the name of the item and a mark for embedding the content of that item.

[0057] For example, for the item "News Summary," the second instruction text 90 includes the text representing the item name "News Summary:" and the mark "@news_summary." If the first prompt 20 includes the news summary, the first prompt generation unit 2040 generates text representing the news summary using the news data 10 and replaces "@news_summary" with the generated text.

[0058] Here, the items in the second instruction text 90 used to generate the first prompt 20 may be all of the items shown in the template 100, or they may be some of the items shown in the template 100. In the latter case, for example, the first prompt generation unit 2040 replaces the marks with text only for the items used in the second instruction text 90 of the template 100. Furthermore, for items where the marks were not replaced with specific text, the first prompt generation unit 2040 deletes the item name and the mark (i.e., the line for that item).

[0059] The items in the second instruction text 90 to be included in the first prompt 20, and the content of those items, are specified, for example, by the user. For example, the first prompt generation unit 2040 presents an input screen that prompts the user for input for each item shown in the second instruction text 90 of the template 100. However, since the text representing the news summary and news keywords is generated using the news data 10, it is not necessary to ask the user for input.

[0060] The content of each item in the second instruction text 90 may be predetermined in correspondence with schedule-related conditions such as dates and days of the week. Hereinafter, the information that corresponds schedule-related conditions with the content of each item in the second instruction text 90 will be referred to as schedule information.

[0061] For example, the schedule information shows combinations of song parameter values ​​for each day of the week. In this case, for example, the first prompt generation unit 2040 obtains the combination of song parameter values ​​corresponding to the current day of the week from the schedule information and includes it in the first prompt 20.

[0062] Figure 6 is an example of schedule information. In Figure 6, the schedule information 200 consists of columns for day of the week 201, genre 202, tempo 203, mood 204, scene 205, and section 206. With this configuration, the schedule information 200 shows the combination of genre, tempo, mood, scene, and section configuration for each day of the week. For example, if the current day of the week is Monday, the first prompt generation unit 2040 obtains the genre, BPM, mood, scene, and section configuration corresponding to Monday from the schedule information 200 and uses them to generate the first prompt 20.

[0063] The method for automatically determining the values ​​of song parameters is not limited to using schedule information 200. For example, the values ​​of song parameters may be defined in association with the type of entity related to the news. The type of entity related to the news may be, for example, the type of organization that caused the incident. In this case, the first prompt generation unit 2040 uses the news data 10 to identify the entity related to the news and includes the values ​​of the song parameters defined in association with the identified entity in the first prompt 20.

[0064] In addition, for example, song parameters may be defined in association with keywords extracted from the news. In this case, the first prompt generation unit 2040 includes the values ​​of the song parameters defined in association with keywords extracted from the news data 10 in the first prompt 20.

[0065] <Generation of lyrics data 30: S106> The lyrics generation unit 2060 generates lyrics data 30 by inputting the first prompt 20 to the first generation model 40 (S106). As mentioned above, the first generation model 40 is configured to output text representing lyrics in response to a prompt instructing the generation of lyrics. Therefore, lyrics data 30 is obtained by inputting the first prompt 20 to the first generation model 40.

[0066] For example, the first generative model 40 is a language model configured to output a response text to a request in response to the input text of the request. For example, the first generative model 40 is a general-purpose language model that has been pre-trained to output a response to any request. In this case, for example, the first generative model 40 is trained using multiple training data sets (hereinafter referred to as the first training data) in which an arbitrary request text is associated with the ground truth response text to that request text. As a language model that outputs a response to any request, for example, any language model classified as an LLM (Large Language Model) can be used. However, the first generative model 40 does not need to be classified as an LLM.

[0067] In addition, for example, the first generative model 40 may be a language model specifically prepared for the lyric generation device 2000. In this case, the first generative model 40 is trained using multiple sets of training data (hereinafter referred to as second training data) in which a first prompt 20 is associated with the ground truth lyrics corresponding to that first prompt 20.

[0068] In addition, the first generative model 40 may also be a language model that has been further trained using the second training data, based on a general-purpose language model generated by training with the first training data.

[0069] <Regarding Lyric Data 30> As mentioned above, the lyrics data 30 includes text data of lyrics based on the content of the news. In addition to the text representing the lyrics, the lyrics data 30 may also include other information related to the lyrics.

[0070] For example, the lyrics data 30 represents the correspondence between sections and lyrics. More specifically, the lyrics data 30 can show the correspondence "section name: lyrics included in that section" for each section.

[0071] Let's assume the section structure is "chorus 1, verse 1, verse 2, chorus 2, and outro". In this case, the lyrics data 30 will show "chorus 1: lyrics for chorus 1", "verse 1: lyrics for verse 1", "verse 2: lyrics for verse 2", and "chorus 2: lyrics for chorus 2".

[0072] The lyrics generation unit 2060 may cause the first generation model 40 to generate the song title. For example, by changing the instruction in the aforementioned first instruction text to the instruction "Create lyrics and title," the first generation model 40 can generate the song title. When a song title is generated in this way, the lyrics data 30 further indicates the said title.

[0073] <Output from Lyrics Generator 2000> The lyrics generation device 2000 can output the processing results in any way. For example, the lyrics generation device 2000 can output lyrics data 30.

[0074] There are various ways in which the lyrics data 30 can be output. For example, the lyrics generation device 2000 stores the lyrics data 30 in an arbitrary memory unit. Alternatively, for example, the lyrics generation device 2000 can output the lyrics data 30 to an arbitrary display device, thereby displaying the contents of the lyrics data 30 on that display device. Alternatively, for example, the lyrics generation device 2000 can transmit the lyrics data 30 to another device (for example, the user terminal mentioned above).

[0075] The lyrics generation device 2000 may also output information other than the lyrics data 30. For example, the lyrics generation device 2000 outputs the first prompt 20 used to generate the lyrics data 30. The output manner of the first prompt 20 is the same as the output manner of the lyrics data 30.

[0076] The lyrics data 30 is output, for example, when the user views the news. In this case, the lyrics generation device 2000 acquires news data 10 representing the content of the news in response to the user viewing the news, and generates the lyrics data 30. The lyrics generation device 2000 then outputs the generated lyrics data 30. This allows the user to view the lyrics generated based on the news along with the news itself. Note that the timing of outputting the lyrics data 30 is not limited to when the user views the news.

[0077] [Embodiment 2] <Overview> Figure 7 is a diagram illustrating an overview of the operation of the lyric generation device 2000. Herein, Figure 7 is a diagram intended to facilitate understanding of the overview of the lyric generation device 2000, and the operation of the lyric generation device 2000 is not limited to what is shown in Figure 7.

[0078] In Embodiment 2, the lyrics generation device 2000 generates song data 70 using the second generation model 60. The song data 70 is data representing the vocal sound of the lyrics represented in the lyrics data 30, or data representing the sound of the vocal sound and the accompanying sound combined.

[0079] The second generation model 60 is configured to output audio data for a song in response to a prompt that indicates an instruction to generate audio data for a song based on specific lyrics. The audio data for a song output from the second generation model 60 represents at least the vocals of the lyrics specified in the prompt. The audio data for a song output from the second generation model 60 may also represent audio that combines the vocals and the accompaniment.

[0080] The lyrics generation device 2000 generates song data 70 based on the lyrics data 30 by inputting a second prompt 50 to the second generation model 60. The second prompt 50 is a prompt that indicates an instruction to generate a song using the lyrics represented by the lyrics data 30. By inputting the second prompt 50 to the second generation model 60, the song data 70 is output from the second generation model 60.

[0081] <Examples of effects and benefits> According to the lyrics generation device 2000, song data 70 containing vocal audio of lyrics generated based on a news summary, news keywords, or both is obtained. By listening to the audio of the song data 70 being played, the user can understand the content of the lyrics without having to read the lyrics shown in the lyrics data 30 themselves. Therefore, the user can understand the content of the lyrics data 30 more easily.

[0082] The lyrics generation device 2000 of this embodiment will be described in more detail below.

[0083] <Example of functional configuration> Figure 8 is a block diagram illustrating the functional configuration of the lyrics generation device 2000. The lyrics generation device 2000 includes an acquisition unit 2020, a first prompt generation unit 2040, and a lyrics generation unit 2060, as well as a music generation unit 2080. The music generation unit 2080 generates music data 70 based on the lyrics data 30 by inputting a second prompt 50 to a second generation model 60.

[0084] <Example of hardware configuration> The hardware configuration of the lyrics generation device 2000 in Embodiment 2 is similar to that of the lyrics generation device 2000 in Embodiment 1, as shown in Figure 3, for example. However, the storage device 1080 in Embodiment 2 stores a program that implements the functions of the lyrics generation device 2000 in Embodiment 2.

[0085] <Processing flow> Figure 9 is a flowchart illustrating the processing flow performed by the lyrics generation device 2000. Steps S102 to S106 are as described with reference to Figure 4. After the execution of S106, the music generation unit 2080 generates music data 70 based on the lyrics data 30 by inputting the second prompt 50 to the second generation model 60 (S202).

[0086] <Regarding the generation of music data 70> The music generation unit 2080 generates music data 70 by inputting the second prompt 50 to the second generation model 60. The second prompt 50 is a prompt that indicates an instruction to generate music using the lyrics represented by the lyrics data 30. For example, the second prompt 50 indicates a standard instruction such as, "Please create music using the lyrics entered." If the lyrics data 30 is not included in the second prompt 50, the music generation unit 2080 inputs the lyrics data 30 along with the second prompt 50 to the second generation model 60. The second generation model 60 generates music data 70 using the lyrics data 30 according to the instructions indicated in the second prompt 50.

[0087] The music generation unit 2080 may generate a second prompt 50 that includes the contents of the lyrics data 30. In this case, for example, the second prompt 50 is generated to show the lyrics shown in the lyrics data 30, following the instruction, "Please create a song using the following lyrics."

[0088] The second prompt 50 may include one or more music parameters. For example, the second prompt 50 may include the same music parameter as the one shown in the first prompt 20. When the same music parameter is included in both the first prompt 20 and the second prompt 50, the value of that music parameter is set to the same value in both the first prompt 20 and the second prompt 50.

[0089] The second prompt 50 may include song parameters not included in the first prompt 20. For example, the lyrics generator 2000 includes the section structure in the first prompt 20, while including genre, tempo, mood, and scene in the second prompt 50.

[0090] The second prompt 50 is generated, for example, using a template, similar to the generation of the first prompt 20. However, the method for generating the second prompt 50 is not limited to using a template.

[0091] The music generation unit 2080 may determine the values ​​of the music parameters to be included in the second prompt 50 in the same manner as the method used to determine the values ​​of the music parameters to be included in the first prompt 20. For example, the music generation unit 2080 may use the schedule information 200 described above to determine the values ​​of the music parameters to be included in the second prompt 50.

[0092] For example, as mentioned above, suppose the section structure is included in the first prompt 20, while the genre, tempo, mood, and scene are included in the second prompt 50. In this case, the first prompt generation unit 2040 refers to the schedule information 200 to identify the genre, tempo, mood, and scene to be included in the first prompt 20. The music generation unit 2080 also refers to the schedule information 200 to identify the section structure to be included in the second prompt 50.

[0093] The method for automatically determining the values ​​of song parameters is not limited to using schedule information 200. For example, the song generation unit 2080 uses news data 10 to identify entities related to the news and includes the values ​​of song parameters defined in association with the identified entities in the second prompt 50. Alternatively, the song generation unit 2080 may also include the values ​​of song parameters defined in association with keywords extracted from the news data 10 in the second prompt 50.

[0094] Here, the lyrics generation device 2000 may generate song data 70 for each of multiple users from a single common lyrics data 30. For example, there are cases where the same news data 10 is used for multiple users, such as when news data 10 is obtained using news rankings. In such cases, the lyrics generation device 2000 can generate song data 70 for each of multiple users using a single lyrics data 30 generated from a single news data 10.

[0095] For example, suppose the values ​​of the song parameters to be included in the second prompt 50 are identified using information provided for each user (for example, schedule information 200 provided for each user). In this case, the content of the second prompt 50 will differ for each user. Therefore, the lyrics generation device 2000 uses the common lyrics data 30 and the second prompt 50 generated for each of the multiple users to generate song data 70 for each of the multiple users.

[0096] <Regarding the second generative model 60> The second generative model 60 may be a general-purpose generative model or a generative model specifically prepared for the lyric generation device 2000. In the former case, the second generative model 60 is a generative model that has been pre-trained to output a song corresponding to any given lyrics. For example, the second generative model 60 is trained using multiple training data sets (hereinafter referred to as the third training data) in which text of arbitrary lyrics is associated with a Ground Truth song corresponding to that text.

[0097] If the second generative model 60 is a generative model specifically prepared for the lyric generation device 2000, the second generative model 60 is trained using multiple sets of training data (hereinafter referred to as the fourth set of training data) in which a second prompt 50 is associated with a Ground Truth song corresponding to that second prompt 50.

[0098] In addition, the second generative model 60 may also be a generative model that has been further trained using a fourth set of training data, in addition to a general-purpose generative model that was generated by training using the third set of training data.

[0099] <Output from Lyrics Generator 2000> The lyrics generation device 2000 can output the processing results in any way. For example, the lyrics generation device 2000 can output song data 70. The output manner of the song data 70 is, for example, the same as the output manner of the lyrics data 30. Alternatively, the lyrics generation device 2000 may output the song represented by the song data 70 as sound by playing the song data 70 and outputting it to any speaker. In addition, the lyrics generation device 2000 may output the sound of the song data 70 to a speaker and display the contents of the lyrics data 30 on a display device.

[0100] The song data 70 is output, for example, when the user views the news. In this case, the lyrics generation device 2000 acquires news data 10 representing the content of the news in response to the user viewing the news, and generates lyrics data 30. Furthermore, the lyrics generation device 2000 uses the lyrics data 30 to generate song data 70. Then, the lyrics generation device 2000 outputs the generated song data 70. This allows the user to listen to a song generated based on the news while viewing the news. Note that the timing of outputting the song data 70 is not limited to when the user views the news.

[0101] Although the present disclosure has been described above with reference to embodiments, the present disclosure is not limited to the embodiments described above. Various modifications to the structure and details of the present disclosure can be made as can be understood by those skilled in the art within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0102] Each drawing is merely illustrative to illustrate one or more embodiments. Each drawing may be associated with one or more other embodiments rather than with only one specific embodiment. As those skilled in the art will understand, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings, for example, to create embodiments not explicitly shown or described. Not all features or steps shown in any one drawing to illustrate an exemplary embodiment are necessarily required, and some features or steps may be omitted. The order of steps shown in any of the drawings may be changed as appropriate.

[0103] Some or all of the embodiments described above may also be described as follows, but are not limited to the following. Some or all of the elements described in any of the following appendices may be applied to various hardware, software, recording means, systems, and methods for recording software.

[0104] Some or all of the above embodiments may also be described as follows, but are not limited to the following: (Note 1) A means of obtaining news data that represents the content of the news, A first prompt generation means generates a first prompt using the news data that instructs the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof. A lyric generation device having a lyric generation means that generates lyric data representing the lyrics by inputting the first prompt into a first generation model. (Note 2) The first prompt generation means generates the first prompt, which indicates the value of one or more parameters relating to the music. The aforementioned parameters are genre, tempo, mood, scene, or section configuration, as described in Appendix 1 of the lyric generation device. (Note 3) The lyrics generation device according to Appendix 2, wherein the first prompt generation means generates the first prompt indicating one or more values ​​of the parameters using schedule information indicating one or more values ​​of the parameters in association with each of a plurality of conditions of the schedule. (Note 4) The lyrics generation device according to Appendix 1, wherein the first prompt generation means generates the first prompt indicating a word or phrase that should not be included in the lyrics, or the type of word or phrase that should not be included in the lyrics. (Note 5) The lyrics generation device described in Appendix 1, wherein the first prompt generation means generates the first prompt indicating the intention to create the lyrics. (Note 6) The lyrics generation device according to Appendix 1, comprising a song generation means for generating song data representing the song by inputting a second prompt to a second generation model that instructs the generation of the song based on the lyrics data. (Note 7) The music generation means generates the second prompt, which indicates the value of one or more parameters relating to the music. The aforementioned parameters are genre, tempo, mood, scene, or section configuration, as described in Appendix 6 of the lyric generation device. (Note 8) The song generation means generates a second prompt indicating one or more values ​​of the parameters using schedule information indicating one or more values ​​of the parameters in correspondence with each of a plurality of conditions of the schedule, as described in Appendix 7. (Note 9) The acquisition step involves obtaining news data that represents the content of the news, A first prompt generation step, which generates a first prompt using the news data, instructing the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof; A lyric generation method performed by a computer, comprising: a lyric generation step of generating lyric data representing the lyrics by inputting the first prompt into a first generation model. (Note 10) The acquisition step involves obtaining news data that represents the content of the news, A first prompt generation step, which generates a first prompt using the news data, instructing the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof; A program that causes a computer to perform a lyric generation step, which generates lyric data representing the lyrics by inputting the first prompt into a first generation model. [Explanation of symbols]

[0105] 10 News Data 20 First Prompt 30 Lyrics Data 40. First Generative Model 50 Second prompt 60 Second Generative Model 70 Song Data 80 First Instruction Text 90 Second Instruction Text 100 templates 200 Schedule Information 201 Day of the week 202 Genres 203 Tempo 204 Mood 205 Scenes Section 206 1000 computers 1020 Bus 1040 processor 1060 memory 1080 Storage Devices 1100 Input / Output Interface 1120 Network Interface 2000 Lyrics generator 2020 Acquisition Department 2040 First prompt generation unit 2060 Lyrics Generation Unit 2080 Music Generation Unit

Claims

1. A means of obtaining news data that represents the content of the news, A first prompt generation means generates a first prompt using the news data that instructs the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof. A lyric generation device having lyric generation means that generates lyric data representing the lyrics by inputting the first prompt into a first generation model.

2. The first prompt generation means generates the first prompt, which indicates the value of one or more parameters relating to the music. The lyric generation device according to claim 1, wherein the parameter is genre, tempo, mood, scene, or section configuration.

3. The lyrics generation device according to claim 2, wherein the first prompt generation means generates the first prompt indicating one or more values ​​of the parameters using schedule information indicating values ​​of one or more of the parameters in association with each of a plurality of conditions of the schedule.

4. The lyrics generation device according to claim 1, wherein the first prompt generation means generates the first prompt indicating a word or phrase that should not be included in the lyrics, or the type of word or phrase that should not be included in the lyrics.

5. The lyrics generation apparatus according to claim 1, wherein the first prompt generation means generates the first prompt indicating the intention to create the lyrics.

6. The lyrics generation device according to claim 1, further comprising a song generation means for generating song data representing the song by inputting a second prompt to a second generation model that instructs the generation of the song based on the lyrics data.

7. The music generation means generates the second prompt, which indicates the value of one or more parameters relating to the music. The lyric generation device according to claim 6, wherein the parameter is genre, tempo, mood, scene, or section configuration.

8. The lyrics generation device according to claim 7, wherein the music generation means generates a second prompt indicating one or more values ​​of the parameters using schedule information indicating one or more values ​​of the parameters in correspondence with each of a plurality of conditions of the schedule.

9. The acquisition step involves obtaining news data that represents the content of the news, A first prompt generation step, which generates a first prompt using the news data, instructing the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof; A lyric generation method performed by a computer, comprising: a lyric generation step of generating lyric data representing the lyrics by inputting the first prompt into a first generation model.

10. The acquisition step involves obtaining news data that represents the content of the news, A first prompt generation step, which generates a first prompt using the news data, instructing the generation of song lyrics based on a summary of the news, keywords of the news, or both thereof; A program that causes a computer to perform a lyric generation step, which generates lyric data representing the lyrics by inputting the first prompt into a first generation model.

Images