JP7474295B2 - Information processing system, information processing method, and program - Google Patents

Description

The present invention relates to an information processing system, an information processing method, and a program.

Patent document 1 discloses a technology that extracts keywords and high-frequency words from text data generated from voice data, and generates minutes using the extracted keywords and high-frequency words.

JP 2022-13586 A

However, the technology described in Patent Document 1 requires tasks such as converting spoken language into written language and formatting the minutes in a format that is suitable for use as minutes, which is problematic in that it requires a lot of man-hours.
In addition, minutes are generated by machine learning, but in this case, it is necessary to learn a large number of pairs of input text data and output sentences corresponding to the input text data, and it is difficult to use highly confidential sentences as training data. Therefore, it is difficult to generate minutes that use a lot of technical terms.
Thus, there has been a problem in that minutes cannot be efficiently generated from text data generated by speech recognition.

The present invention has been made in consideration of the above points, and aims to provide an information processing system, information processing method, and program that can efficiently generate minutes from text data generated by voice recognition.

The present invention has been made to solve the above problems, and one aspect of the present invention is an information processing system including a keyword acquisition unit that acquires keywords extracted from text data by speech recognition of voice data, and a first sentence generation unit that generates a first sentence including the keywords.

Another aspect of the present invention is an information processing method including a keyword acquisition process in which a computer of an information processing device acquires keywords extracted from text data by speech recognition of speech data, and a first sentence generation process in which a first sentence including the keywords is generated.

Another aspect of the present invention is a program for causing a computer of an information processing device to execute a keyword acquisition step of acquiring keywords extracted from text data by speech recognition of voice data, and a first sentence generation step of generating a first sentence including the keywords.

According to the present invention, minutes can be efficiently generated from text data generated by speech recognition.

1 is a system configuration diagram showing an example of a configuration of an information processing system according to a first embodiment of the present invention. 1 is a block diagram showing an example of a configuration of a conversion device according to a first embodiment of the present invention. 1 is a block diagram showing an example of a configuration of a keyword extraction device according to a first embodiment of the present invention. 1 is a block diagram showing an example of a functional configuration of an information processing apparatus according to a first embodiment of the present invention; 1 is a block diagram showing an example of a hardware configuration of an information processing apparatus according to a first embodiment of the present invention. 4 is a flowchart showing an example of information processing in the information processing system according to the first embodiment of the present invention. FIG. 2 is a diagram showing an example of a display screen in the information processing apparatus according to the first embodiment of the present invention.

First Embodiment
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.
<Configuration of Information Processing System>
First, the configuration of the information processing system will be described.
FIG. 1 is a system configuration diagram showing an example of the configuration of an information processing system according to a first embodiment of the present invention.
The information processing system includes a conversion device 1, a keyword extraction device 2, and an information processing device 3. The conversion device 1, the keyword extraction device 2, and the information processing device 3 are communicatively connected to each other via a network NW.

The conversion device 1 has a function of converting voice data into text data. When voice data is input, the conversion device 1 converts the voice data into text data by voice recognition. The conversion device 1 transmits the converted text data to the keyword extraction device 2 via the network NW.

The keyword extraction device 2 has a function of extracting keywords from text data. When the keyword extraction device 2 receives text data from the conversion device 1, it extracts a predetermined number of keywords from the text data. The keyword extraction device 2 transmits the extracted predetermined number of keywords as keyword candidates to the information processing device 3 via the network NW.

The information processing device 3 has a function of generating one or more sentences including a keyword selected from the keyword candidates. The one or more sentences including the keyword include all of the one or more keywords selected from the keyword candidates. Here, in the following explanation, the multiple sentences including the keyword are referred to as first sentence candidates. Also, in the following explanation, one sentence selected from the first sentence candidates is referred to as the first sentence.

Note that if the first sentence is only one sentence, that one sentence is both a candidate for the first document and the first sentence.

The first sentence candidates and the first sentences are generated for each predetermined amount of text data. The predetermined amount may be, for example, for each sentence, for multiple sentences, for each piece of text data corresponding to a predetermined time of audio data, or for each predetermined number of characters of text data.

The predetermined amount may be set by storing setting information of the predetermined amount in advance in the information processing device 3, or may be set by the user as a sentence break setting when starting information processing, which will be described later, and the information processing device 3 may store the setting information of the predetermined amount. The setting information of the predetermined amount is information for setting the predetermined amount.
When the keyword is changed or the first sentence is selected, the information processing device 3 generates a candidate for the first sentence including the new keyword. Here, the first sentence is a template sentence that serves as a template for generating the second sentence, for example, minutes. The template sentence is a sentence generated by machine learning when the selected keyword is input.

The information processing device 3 generates a second sentence that includes a first sentence for each selected keyword. In other words, the information processing device 3 generates a second sentence by combining a plurality of first sentences. The second sentence can be edited by the user.

In this manner, the information processing system according to this embodiment extracts keyword candidates from text data generated by speech recognition. The information processing system generates a first sentence candidate that includes all keywords selected from the keyword candidates. The information processing system generates an editable second sentence that includes a first sentence selected from the first sentence candidates generated for each selected keyword.

This allows the information processing system to generate a template first sentence based on the learning results from machine learning and the selected keywords, thereby efficiently generating minutes.

<Configuration of conversion device 1>
Next, the conversion device 1 will be described.
FIG. 2 is a block diagram showing an example of the configuration of the conversion device 1 according to the first embodiment of the present invention.
The conversion device 1 includes a communication unit 11, a control unit 12, and a storage unit 13.

<Communication unit 11>
The communication unit 11 has a function of communicating with other devices via a network. The communication unit 11 outputs data received from other devices to the control unit 12. The communication unit 11 also transmits data input from the control unit 12 to other devices.
<Storage unit 13>
The storage unit 13 has a function of storing various data, applications, and programs.

<Control Unit 12>
The control unit 12 has a function of controlling the conversion device 1. The control unit 12 reads out various data, applications, programs, etc. stored in the storage unit 13 and controls the conversion device 1.

The processing of the control unit 12 will now be described in more detail.
The control unit 12 includes a voice data acquisition unit 121 , a voice recognition unit 122 , and a text data output unit 123 .

<Audio Data Acquisition Unit 121>
The voice data acquisition unit 121 has a function of acquiring voice data. The voice data acquisition unit 121 acquires the voice data via an input unit (not shown). The voice data may be acquired by receiving voice data transmitted from another device via the communication unit 11, or by reading voice data previously stored in the storage unit 13. The voice data acquisition unit 121 outputs the acquired voice data to the voice recognition unit 122.

<Speech Recognition Unit 122>
The voice recognition unit 122 has a function of converting voice data into text data through voice recognition. The voice recognition is performed by a voice recognition engine. The voice recognition unit 122 outputs the converted (generated) text data to the text data output unit 123.

<Text Data Output Unit 123>
The text data output unit 123 has a function of outputting the text data generated by the voice recognition unit 122. The text data output unit 123 transmits the text data to the keyword extraction device 2 via the communication unit 11.

<Keyword Extraction Device 2>
Next, the keyword extractor 2 will be described.
FIG. 3 is a block diagram showing an example of the configuration of the keyword extracting device 2 according to the first embodiment of the present invention.
The keyword extractor 2 includes a communication unit 21, a control unit 22, and a storage unit 23.

<Communication unit 21>
The communication unit 21 has a function of communicating with other devices via a network. The communication unit 21 outputs data received from other devices to the control unit 22. The communication unit 21 also transmits data input from the control unit 22 to other devices.

<Storage unit 23>
The storage unit 23 has a function of storing various data, applications, and programs.

<Control unit 22>
The control unit 22 has a function of controlling the keyword extraction device 2. The control unit 22 reads out various data, applications, programs, etc. stored in the storage unit 23 to control the keyword extraction device 2. The process of the control unit 22 will be described in more detail.
The control unit 22 includes a text data acquisition unit 221 , a setting information acquisition unit 222 , a text extraction unit 223 , a keyword extraction unit 224 , and a keyword output unit 225 .

<Text Data Acquisition Unit 221>
The text data acquisition unit 221 has a function of acquiring text data from the conversion device 1 via the communication unit 21. The text data acquisition unit 221 outputs the acquired text data to the text extraction unit 223.

<Setting information acquisition unit 222>
The setting information acquisition unit 222 has a function of acquiring setting information from the storage unit 23. The setting information is information on a predetermined amount of text data from which keywords are to be extracted, and information on the number of extracted keywords. The information on the number of extracted keywords is information on the number of keywords that the keyword extraction unit 224 extracts from the text data. The setting information acquisition unit 222 reads out the setting information from the storage unit 23. The setting information acquisition unit 222 outputs the read out setting information to the text extraction unit 223 and the keyword extraction unit 224.

<Text Extraction Unit 223>
The text extraction unit 223 has a function of extracting text for each predetermined amount from the text data acquired by the text data acquisition unit 221. The text extraction unit 223 outputs the text data corresponding to the extracted text for each predetermined amount to the keyword extraction unit 224.

<Keyword Extraction Unit 224>
The keyword extraction unit 224 has a function of extracting keywords from the text data input from the text extraction unit 223. The keyword extraction unit 224 extracts the number of keywords indicated by the setting information for each piece of text data. The keyword extraction unit 224 outputs the extracted keywords to the keyword output unit 225.

Here, the keyword extraction unit 224 uses term frequency-inverse document frequency (TF-IDF), PositionRank, bidirectional encoder representations from transformers (BERT), etc. to extract keywords. TF-IDF is a value that increases in proportion to the number of times a word appears or the frequency of its appearance in a sentence, and is a statistic (numerical value) that indicates how important a word is in a corpus or a group of collected documents. PositionRank is an algorithm that calculates the importance of each word and extracts key phrases using words with high importance. BERT is a transformer-based machine learning method for pre-learning of natural language processing. That is, the keyword extraction unit 224 extracts keywords as many as the number of keywords indicated in the setting information for each piece of text data input from the text extraction unit 223, using a keyword extraction means based on the number of times a word appears or the frequency of appearance of a word, such as TF-IDF or PositionRank, or a keyword extraction means based on the importance of a word calculated from the context, such as BERT. Note that the keyword extraction means is an example, and other general keyword extraction means, such as MultipartiteRank, TopicRank, YAKE, SGRank, and sCAKE, may also be used.

<Keyword Output Unit 225>
The keyword output unit 225 has a function of outputting a keyword, and transmits the keyword input from the keyword extraction unit 224 to the information processing device 3 via the communication unit 21.

<Information processing device 3>
Next, the information processing device 3 will be described.
FIG. 4 is a block diagram showing an example of a functional configuration of the information processing device 3 according to the first embodiment of the present invention.
The information processing device 3 includes a communication unit 31, a control unit 32, and a storage unit 33. The control unit 32 includes a keyword acquisition unit 321, a keyword candidate presentation unit 322, a first sentence generation unit 323, a first sentence candidate presentation unit 324, a second sentence generation unit 325, an output unit 326, and a learning unit 327.

<Communication unit 31>
The communication unit 31 has a function of communicating with other devices via a network. The communication unit 31 outputs data received from other devices to the control unit 32. The communication unit 31 also transmits data input from the control unit 32 to other devices.
<Storage unit 33>
The storage unit 33 has a function of storing various data, applications, and programs.

<Control unit 32>
The control unit 32 has a function of controlling the information processing device 3. The control unit 32 reads out various data, applications, programs, etc. stored in the storage unit 33 and controls the information processing device 3.

<Keyword Acquisition Unit 321>
The keyword acquisition unit 321 has a function of acquiring keywords. The keyword acquisition unit 321 acquires keywords received from the keyword extraction device 2 via the communication unit 31. The keyword acquisition unit 321 stores the acquired keywords as keyword candidates in the storage unit 33. The keyword acquisition unit 321 also outputs the acquired keywords to the keyword candidate presentation unit 322 as keyword candidates.

<Keyword Candidate Presentation Unit 322>
The keyword candidate presentation unit 322 has a function of presenting keyword candidates and allowing the user to select a keyword. The keyword candidate presentation unit 322 presents one or more keyword candidates input from the keyword acquisition unit 321 to the user via the output unit 326. When there are multiple keywords, the keyword acquisition unit 321 ranks each keyword in order of importance, frequency, or frequency of appearance of the keyword according to the importance of the keyword at the time of keyword extraction by the keyword extraction device 2, the frequency of appearance of the keyword, the number of times the keyword appears, and the like, and presents a predetermined number of keywords in order from the highest to lowest ranking. The predetermined number may be a predetermined number of keywords, a number that can be presented as keyword candidates, or all keyword candidates.

The keyword candidate presentation unit 322 accepts a keyword selection operation by the user based on the presentation of keyword candidates. The keyword candidate presentation unit 322 outputs the selected keyword candidates to the first sentence generation unit 323 as selected keywords based on the user's selection of one or more keywords from the presented keyword candidates. The keyword candidate presentation unit 322 outputs the selected keywords to the first sentence generation unit 323 based on the user's change of the selected keywords.

In addition to the keyword candidates, the keyword candidate presentation unit 322 may present any keyword input by the user as a keyword candidate.
In this way, it is possible to select and change selected keywords from keyword candidates including arbitrary keywords, so that expressions that align with the user's intention can be used, not limited to words that appear in utterances (text data) as in the past. Therefore, minutes can be generated efficiently.

<First Sentence Generation Unit 323>
The first sentence generation unit 323 has a function of generating a first sentence including a selected keyword based on the input selected keyword. The first sentence generation unit 323 generates a first sentence including the selected keyword based on the selected keyword input from the keyword candidate presentation unit 322. A first sentence including a selected keyword is a sentence including all of the selected keywords. The first sentence generation unit 323 generates multiple different first sentences using the same selected keyword. The first sentence generation unit 323 outputs the generated first sentence to the first sentence candidate presentation unit 324 as a first sentence candidate.

Here, the first sentence is generated by using a generative Transformer model to generate the first sentence including the selected keyword. The generative Transformer model is, for example, a Generative Pretrained Transformer (GPT). GPT is an autoregressive language model that automatically generates sentences using deep learning called a Transformer. GPT generates the first sentence by repeatedly predicting the word that follows a certain word based on the occurrence probability distribution in the training data. GPT uses a learning model that takes a keyword as input and the first sentence as output. The learning model is one or both of a general-purpose learning model called a base model and a dedicated learning model called fine tuning (also called a customized model). The customized model is a dedicated learning model suitable for a customer, industry, department, field, theme, etc.

In this way, whereas in the past it was necessary to train using a large number of pairs of input text data generated by speech recognition and output minutes, it is now possible to train using pairs of input keywords and output first sentences, allowing for efficient machine learning. Furthermore, when training using a large number of pairs, it was difficult to use minutes containing confidential information, but since it is only necessary to train pairs of keywords and first sentences, it is possible to train using a dedicated training model. As a result, it is possible to generate first sentences suitable for the field desired by the user.

<First Sentence Candidate Presentation Unit 324>
The first sentence candidate presentation unit 324 has a function of presenting the first sentence candidates and, as a result, accepting a selection operation of the first sentence. The first sentence candidate presentation unit 324 presents the first sentence candidates input from the first sentence generation unit 323 to the user via the output unit 326. The first sentence candidate presentation unit 324 outputs the selected first sentence to the second sentence generation unit 325 based on the user's selection of the first sentence from the first sentence candidates. In addition, the first sentence candidate presentation unit 324 stores the selected first sentence in the storage unit 33 in association with a keyword included in the first sentence.

<Second Sentence Generation Unit 325>
The second sentence generation unit 325 has a function of generating a second sentence including the selected first sentence by the selected first sentence. The second sentence is a sentence that includes the first sentence selected for each predetermined amount of text data and is generated from a group of first sentences. The second sentence generation unit 325 outputs the second sentence to the output unit 326.

<Output unit 326>
The output unit 326 has a function of outputting various information output from the keyword candidate presentation unit 322, the first sentence candidate presentation unit 324, and the second sentence generation unit 325. The output unit 326 includes a display device, and causes the display device to display the keyword candidates, the first sentence candidates, the second sentences, etc.

<Learning Unit 327>
The learning unit 327 has a function of learning a learning model for generating a first sentence. The learning model used by the learning unit 327 is assumed to have been learned in advance using a base model and a customized model. The learning model is a learning model that receives a keyword as input and obtains a first sentence as output. The learning unit 327 reads out the keyword stored in the memory unit 33 and the first sentence containing the keyword, and performs reinforcement learning.

<Hardware Configuration>
Next, a description will be given of the hardware configuration of the information processing device 3. Note that, although a description of the hardware configurations of the conversion device 1 and the keyword extraction device 2 will be omitted, they have the same configuration as that shown in the block diagram of FIG.
FIG. 5 is a block diagram showing an example of a hardware configuration of the information processing device 3 according to the first embodiment of the present invention.
The information processing device 3 includes a CPU 301, a storage medium interface unit 302, a storage medium 303, an input device 304, an output device 305, a ROM 306 (Read Only Memory), a RAM 307 (Random Access Memory), an auxiliary storage unit 308, and a network interface unit 309. The CPU 301, the storage medium interface unit 302, the input device 304, the output device 305, the ROM 306, the RAM 307, the auxiliary storage unit 308, and the network interface unit 309 are connected to each other via a bus.
The CPU 301 referred to here refers to a processor in general, and includes not only a device called a CPU in the narrow sense, but also, for example, a GPU, a DSP, etc. The CPU 301 referred to here is not limited to being realized by one processor, and may be realized by combining multiple processors of the same or different types.

<CPU 301>
The CPU 301 controls the information processing device 3 by reading and executing programs stored in the auxiliary storage unit 308, ROM 306, and RAM 307, and by reading various data stored in the auxiliary storage unit 308, ROM 306, and RAM 307 and writing the various data to the auxiliary storage unit 308 and RAM 307. The CPU 301 also reads various data stored in the storage medium 303 via the storage medium interface unit 302, and writes the various data to the storage medium 303.
<Storage medium 303>
The storage medium 303 is a portable storage medium such as a magneto-optical disk, a flexible disk, or a flash memory, and stores various data.

<Storage medium interface unit 302>
The storage medium interface unit 302 is an interface for reading and writing data from and to the storage medium 303 .

<Input Device 304>
The input device 304 is an input device such as a mouse, a keyboard, a touch panel, a volume control button, a power button, a setting button, and an infrared receiver.

<Output Device 305>
The output device 305 is an output device such as a display unit and a speaker.

<ROM 306, RAM 307>
The ROM 306 and the RAM 307 store programs and various data for operating each functional unit of the conversion device 1 .

<Auxiliary Storage Unit 308>
The auxiliary storage unit 308 is a hard disk drive, a flash memory, or the like, and stores programs for operating each functional unit of the information processing device 3 and various data.

<Network Interface Unit 309>
The network interface unit 309 has a communication interface and is connected to the network NW via wireless communication.

For example, in the functional configuration of the information processing device 3, the control unit 32 corresponds to the CPU 301, the memory unit 33 corresponds to the ROM 306, the RAM 307, the auxiliary memory unit 308, or any combination thereof, the output unit 326 corresponds to the output device 305, and the communication unit 31 corresponds to the network interface unit 309.

<Flowchart>
Next, an example of information processing by the information processing system will be described.
FIG. 6 is a flowchart showing an example of information processing in the information processing system according to the first embodiment of the present invention.
In step S100, the conversion device 1 converts the voice data into text data by speech recognition. The conversion device 1 transmits the converted text data to the keyword extraction device 2.
In step S102, the keyword extractor 2 extracts keywords from a predetermined amount of text data. The keyword extractor 2 transmits the extracted keywords to the information processing device 3.

In step S104, the information processing device 3 acquires the keywords extracted by the keyword extracting device 2 as keyword candidates.
In step S106, the information processing device 3 presents the keyword candidates to the user via the output unit 326.

In step S108, the information processing device 3 accepts a keyword selected by a user operation from among the presented keyword candidates.
In step S110, the information processing device 3 determines whether or not the keyword selection operation by the user is completed. If the keyword selection operation is completed (step S110: YES), the information processing device 3 executes the process of step S112. On the other hand, if the keyword selection operation is not completed (step S110: NO), the information processing device 3 executes the process of step S108. Here, whether or not the selection operation is completed may be determined based on the passage of a predetermined time after the operation signal by the user operation is no longer input, or may be determined based on the fact that the processing after the keyword selection is being executed.

In step S112, the information processing device 3 generates a plurality of types of first sentences including all of the selected keywords.
In step S114, the information processing device 3 presents the generated multiple types of first sentences as first sentence candidates to the user via the output unit 326.
In step S116, the information processing device 3 generates a second sentence from the first sentence selected by the user from the first sentence candidates. The keyword extraction device 2 repeats the process from step S102 to extract keyword candidates for the next predetermined amount of text data, and the information processing device 3 then repeatedly executes the process from step S104 to step S116 for the keyword candidates for the next predetermined amount of text data. When the information processing device 3 generates a second sentence including the last first sentence for the text data, it ends the process related to FIG. 6.

In addition, in step S108, the information processing device 3 may allow the user to input an arbitrary keyword through a user operation, in addition to or instead of selecting a keyword from the presented keyword candidates through a user operation, and may use the arbitrary keyword as the selected keyword.

<Display screen example>
Next, examples of display screens will be described.
FIG. 7 is an example of a display screen displayed on the output unit of the information processing apparatus according to the first embodiment of the present invention.
The illustrated display screen example is a two-screen display example consisting of a first screen G1 and a second screen G2. The illustrated example is an example of a case where a first sentence is generated using keywords extracted from a predetermined amount of text data, and a portion B of the second sentence is generated by the selected first sentence.
The first screen G1 displays the heading word KWC "Keyword candidate" which indicates a keyword candidate.
Each of the keyword candidates is displayed in a display window w1. As an example, a keyword KW1 "machine learning" and a keyword KW2 "constraint" are displayed as keyword candidates.

To increase the number of keyword candidates, the keyword candidates displayed in the display window w1 can be increased by checking the check box CB "Increase keywords." The example shown in the figure shows an example of a case where the keyword KW3 "Data collection" is additionally displayed as a keyword candidate in the display window w1 based on checking the check box CB.

The first screen G1 also displays the entry word KWS "selected keyword", which indicates a keyword selected from the keyword candidates. The selected keyword is displayed in a display window w2. For example, to select a keyword from the keyword candidates, each keyword displayed in the display window w1 is dragged and dropped into the display window w2. The keyword moved to the display window w2 indicates that it is the selected keyword.
In the illustrated example, the keyword KW1 "machine learning" and the keyword KW2 "constraint" are selected as the selected keywords.

Further, the first screen G1 displays the entry word SNC "first sentence candidate" representing the first sentence candidate. The first sentence candidates are displayed as items SN1, SN2, and SN3, respectively.
In item SN1, the first sentence candidates are displayed as follows: "Candidate A,""PoC using machine learning is being implemented. One concern is that the man-hours required for data collection are running short..." The first sentences generated based on the selected keywords are displayed.
Item SN2 displays another first sentence generated based on the selected keywords as a first sentence candidate, such as "Candidate B" and "Employee workload is tight, so we will improve efficiency by introducing an application that utilizes machine learning...".
Here, "candidate A" and "candidate B" are different first sentences generated by the same keyword as described above.

The second screen G2 displays the entry word SNS "Selected First Sentence" representing the selected first sentence.
To select a first sentence from among the first sentence candidates, items SN1, SN2, SN3, etc., representing the first sentence candidates respectively, are dragged and dropped into area SN.
As an example, in area SN, the following is displayed as the selected first sentence: "Candidate A""PoC using machine learning is being implemented. One concern is that the man-hours required for data collection are running short..."

Further, the second screen G2 displays the entry word SND "Second Sentence" representing the second sentence.
The second sentences generated by the respective first sentences are displayed in a display window w3.
In the illustrated example, the selected first sentence is displayed in part B SS2 in the display window w3, "PoC utilizing machine learning is being implemented. One concern is that the man-hours required for data collection are running short..."

In addition, in portion A "..." SS1, a first sentence is displayed that is generated from a predetermined amount of text data in a portion of the text data preceding the text data corresponding to the keyword in portion B. In addition, in portion C, a first sentence is displayed that is based on a keyword extracted from a predetermined amount of text data in the next portion of the text data.
Here, in the second sentence, it is possible to change the order of the first sentence, for example, part A, part B, and part C. For example, it is possible to edit the second sentence, for example, by changing the order to part C, part B, and part A.

Note that an arbitrary keyword may be input by a user operation, and the arbitrary keyword may be used as a selected keyword. In this case, the information processing device 3 may display, for example, in the display window w2, the arbitrary keyword input by the user as a selected keyword in addition to the keyword KW1 "machine learning" and the keyword KW2 "pressure", and generate a first sentence that includes all selected keywords including the arbitrary keyword.

In this way, the information processing device 3 according to this embodiment includes a keyword acquisition unit 321 that acquires keywords extracted from text data by voice recognition of voice data, and a first sentence generation unit 323 that generates a first sentence including the keywords.

This allows machine learning to generate sentences using keywords extracted from the text data generated by voice recognition, making it possible to generate minutes efficiently.

The information processing device 3 further includes a first sentence candidate presentation unit 324 that presents multiple candidates for the generated first sentence, and a second sentence generation unit 325 that generates a second sentence including the first sentence based on the result of the user's selection of the first sentence.

The user can select the intended sentence from multiple types of first sentences generated from the same keyword, so the second sentence can be generated efficiently. As a result, minutes can be generated efficiently.

The information processing device 3 further includes an output unit 326 that outputs the generated second sentence so that it can be edited by the user.

This improves user convenience and the quality of the second sentence.

The information processing device 3 further includes a keyword candidate presentation unit 322 that presents one or more keywords extracted from the text data, and a first sentence generation unit 323 that generates one or more first sentences that include all of the one or more keywords selected by the user.

In this way, a first sentence can be generated using keywords intended by the user, and therefore a second sentence can be generated efficiently. Furthermore, even if keywords extracted from text data are in colloquial language, they can be changed to keywords in formal language, and therefore a second sentence can be generated efficiently. Furthermore, user convenience can be improved.

The first sentence generation unit 323 also uses the acquired one or more keywords as input to generate a first sentence through machine learning using a learning model based on at least one of a generic model and a customized model.

This improves learning efficiency because it is necessary to learn only pairs of keywords and sentences that contain those keywords.

Furthermore, the keyword acquiring unit 321 acquires keywords from each predetermined amount of text data, and the first sentence generating unit 323 generates a first sentence including the keywords for each predetermined amount of text data.
This allows the information processing system to generate the first sentence by acquiring keywords corresponding to a range of text data of a predetermined amount, thereby enabling the generation of the first sentence as intended by the user. Also, the generation accuracy of the first sentence as intended by the user can be improved.

The above describes in detail the embodiments of the present invention with reference to the drawings, but the specific configuration is not limited to the above, and various design changes can be made without departing from the spirit of the present invention.

For example, in the above embodiment, an example was described in which the conversion device 1, the keyword extraction device 2, and the information processing device 3 are configured as individual devices, but one aspect of the present invention may be realized by a device that combines some or all of these devices.

The programs that run on the conversion device 1, keyword extraction device 2, and information processing device 3 in one aspect of the present invention may be programs that control one or more processors such as a CPU (Central Processing Unit) (programs that cause a computer to function) so as to realize the functions shown in the above-mentioned embodiments and modified examples related to one aspect of the present invention. Information handled by each of these devices may be temporarily stored in a RAM (Random Access Memory) during processing, and then stored in various storage devices such as a flash memory or HDD (Hard Disk Drive), and may be read, modified, and written by the CPU, etc. as necessary.

In addition, part or all of the conversion device 1, keyword extraction device 2, and information processing device 3 in each of the above-mentioned embodiments and variations may be realized by a computer having one or more processors. In that case, the control function may be realized by recording a program for realizing the control function on a computer-readable recording medium, and reading and executing the program recorded on the recording medium into a computer system.

The "computer system" referred to here is the computer system built into the conversion device 1, keyword extraction device 2, and information processing device 3, and includes hardware such as the OS and peripheral devices. Also, the "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into the computer system.

Furthermore, "computer-readable recording medium" may also include something that dynamically holds a program for a short period of time, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line, or something that holds a program for a certain period of time, such as volatile memory inside a computer system that serves as a server or client in such a case. The above program may also be one that realizes part of the functions described above, or one that can realize the functions described above in combination with a program already recorded in the computer system.

In addition, the conversion device 1, keyword extraction device 2, and information processing device 3 in each of the above-mentioned embodiments and modifications may be realized in part or in whole as an LSI, which is typically an integrated circuit, or as a chip set. In addition, each functional block of the conversion device 1, keyword extraction device 2, and information processing device 3 in each of the above-mentioned embodiments and modifications may be individually formed into a chip, or may be integrated in part or in whole into a chip. The integrated circuit method is not limited to LSI, and may be realized by a dedicated circuit and/or a general-purpose processor. In addition, if an integrated circuit technology that can replace LSI appears due to advances in semiconductor technology, it is also possible to use an integrated circuit based on that technology.

Although the above describes in detail each embodiment and modification as one aspect of this invention with reference to the drawings, the specific configuration is not limited to each embodiment and modification, and also includes design changes within the scope of the gist of this invention. Furthermore, various modifications of one aspect of the present invention are possible within the scope of the claims, and embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. Also included are configurations in which elements described in the above embodiments and modifications are substituted with elements that have the same effect.

REFERENCE SIGNS LIST 1 Conversion device 11 Communication unit 12 Control unit 121 Voice data acquisition unit 122 Voice recognition unit 123 Text data output unit 13 Memory unit 2 Keyword extraction device 21 Communication unit 22 Control unit 221 Text data acquisition unit 222 Setting information acquisition unit 223 Text extraction unit 224 Keyword extraction unit 225 Keyword output unit 23 Memory unit 3 Information processing device 31 Communication unit 32 Control unit 321 Keyword acquisition unit 322 Keyword candidate presentation unit 323 First sentence generation unit 324 First sentence candidate presentation unit 325 Second sentence generation unit 326 Output unit 327 Learning unit 33 Memory unit 301 CPU
302 storage medium interface unit 303 storage medium 304 input device 305 output device 306 ROM
307 RAM
308 Auxiliary storage unit 309 Network interface unit

Claims (6)

a keyword acquiring unit that acquires keywords extracted from text data by speech recognition of the voice data;
a first sentence generating unit that generates a plurality of first sentences including the keyword;
a first sentence candidate presentation unit that presents each of the generated first sentence candidates;
a second sentence generating unit that generates a second sentence including the first sentence based on a selection result of the first sentence by a user;
Equipped with
The first sentence generation unit generates the first sentences by machine learning using a learning model based on at least one of a base model and a customized model, using the acquired one or more keywords as an input.
Information processing system. An output unit that outputs the generated second sentence in a manner that the user can edit the second sentence is further provided.
The information processing system according to claim 1 . a keyword candidate presentation unit that presents one or more keywords extracted from the text data;
Further equipped with
The first sentence generation unit generates a plurality of the first sentences including all of the one or more keywords selected by the user.
The information processing system according to claim 1 . the keyword acquisition unit acquires the keyword for each predetermined amount of text data;
the first sentence generation unit generates a plurality of the first sentences including the keyword for each of the predetermined amount of text data.
The information processing system according to claim 1 . The computer of the information processing device
a keyword acquisition step of acquiring keywords extracted from text data by speech recognition of the voice data;
a first sentence generating step of generating a plurality of first sentences including the keyword;
a first sentence candidate presenting step of presenting each of the generated first sentence candidates;
a second sentence generating step of generating a second sentence including the first sentence based on a selection result of the first sentence by a user;
having
In the first sentence generation step, the acquired one or more keywords are used as an input to generate a plurality of the first sentences by machine learning using a learning model based on at least one of a base model and a customized model.
Information processing methods. The computer of the information processing device
a keyword acquisition step of acquiring keywords extracted from text data by speech recognition of the speech data;
a first sentence generating step of generating a plurality of first sentences including the keyword;
a first sentence candidate presenting step of presenting each of the generated first sentence candidates;
a second sentence generating step of generating a second sentence including the first sentence based on a selection result of the first sentence by a user;
Run the command,
In the first sentence generating step, the acquired one or more keywords are used as an input to generate the first sentences by machine learning using a learning model based on at least one of a base model and a customized model.
program.

Images