Natural language processing device and computer-readable recording medium

Abstract

A natural language processing device according to the present disclosure comprises: a verification data storage unit that stores at least text data being verified, which is used in processing; a processing data input unit that receives input of text data being processed; an inference unit that acquires an inference result based on the text data being processed; an auxiliary information creation unit that creates auxiliary information including at least a result of having divided the text data being processed by means of a prescribed text division algorithm used in order to obtain the inference result and a result of having divided the text data being verified by means of the prescribed text division algorithm; and an output unit that outputs the inference result and the auxiliary information.

Description

Natural language processing device and computer-readable recording medium

[0001] The present disclosure relates to a natural language processing device and a computer-readable recording medium.

[0002] In recent years, attempts at DX using natural language processing techniques such as generative AI (Artificial Intelligence) and large language models (LLM) have been made in various industrial fields such as manufacturing (see, for example, Patent Literature 1). There is also a movement to apply LLMs and the like to business operations in industrial fields such as factories.

[0003] Japanese Patent Application Laid-Open No. 2020-135457

[0004] In the industrial field, there is a lot of unique terminology and noise (word abbreviations, spelling conventions, etc.) used in each industry and manufacturing site. Therefore, when applying technologies such as LLM in this field, even if general methods are used as is, accuracy may be insufficient. Furthermore, there are many situations where high accuracy is required for frequently used specific type inputs. In particular, in the drive control of industrial machinery, incorrect operation or incorrect setting can result in significant losses. Therefore, when applying these technologies, it is necessary to maintain a certain level of accuracy while also ensuring the explainability of the content output by generative AI and LLM. However, the natural language processing technology used in generative AI and LLM has a more complex inference process than conventional machine learning, making it difficult to explain results based on internal specifications. Therefore, there is a need for technology that can confirm the LLM's business application's interpretation of the query content.

[0005] The natural language processing device disclosed herein solves the above-mentioned problem by displaying various user assistance information on the screen for industrial business applications that use inference processing using LLM, such as displaying the handling of unique words and driving related words in the input sentence text segmentation results, thereby allowing the user to confirm whether the output results are based on internal processing in line with their intention.

[0006] One aspect of the present disclosure is a natural language processing device comprising: a verification data storage unit that stores at least verification target text data to be used in processing; a processing data input unit that accepts input of processing target text data; an inference unit that acquires an inference result based on the processing target text data; an auxiliary information creation unit that creates auxiliary information including at least the result of text segmentation of the processing target text data using a predetermined text segmentation algorithm used to obtain the inference result and the result of text segmentation of the verification target text data using the predetermined text segmentation algorithm; and an output unit that outputs the inference result and the auxiliary information.

[0007] 1 is a schematic hardware configuration diagram of a natural language processing apparatus according to an embodiment of the present disclosure; FIG. 2 is a block diagram showing the schematic functions of a natural language processing apparatus according to a first embodiment; FIG. 3 is a schematic diagram showing data exchanged between a natural language processing apparatus and another apparatus; FIG. 4 is a table diagram showing an example of data related to a business application stored in a business application data storage unit; FIG. 5 is a table diagram showing an example of verification target text data stored in a verification data storage unit; FIG. 6 is a screen diagram showing an example of a connection screen of a business application apparatus via a natural language processing apparatus; FIG. 7 is a screen diagram showing an example in which an inference result by a business application apparatus and auxiliary information created by the natural language processing apparatus are displayed together; FIG. 8 is a block diagram showing the schematic functions of a natural language processing apparatus according to a second embodiment; FIG. 9 is a block diagram showing the schematic functions of a natural language processing apparatus according to another embodiment;

[0008] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following description, components having the same or similar functions will be denoted by the same reference numerals. Duplicate descriptions of those components may be omitted.

[0009] In this application, "based on XX" means "based on at least XX," and includes cases where it is based on other elements in addition to XX. Furthermore, "based on XX" is not limited to cases where XX is used directly, but also includes cases where it is based on XX that has been calculated or processed. "XX" is any element (for example, any information).

[0010] 1 is a schematic hardware configuration diagram showing the main parts of a natural language processing device according to a first embodiment of the present disclosure. The natural language processing device 1 according to this embodiment can be implemented on a computer such as a personal computer attached to a control device, a personal computer connected to the control device via a wired / wireless network, a cell computer, a fog computer 6, or a cloud server 7. This embodiment shows an example in which the natural language processing device 1 is implemented on a computer connected via a network to a control device 3 that controls industrial machinery 4, and a business application device 2 that uses generation AI, LLM, etc.

[0011] The CPU 11 provided in the natural language processing apparatus 1 according to this embodiment is a processor that controls the entire natural language processing apparatus 1. The CPU 11 reads a system program stored in the ROM 12 via the bus 22, and controls the entire natural language processing apparatus 1 in accordance with the system program. The RAM 13 temporarily stores temporary calculation data, display data, various data acquired from the outside, and the like.

[0012] The non-volatile memory 14 is composed of, for example, a battery-backed memory (not shown) or an SSD (Solid State Drive), and retains its stored state even when the natural language processing apparatus 1 is powered off. The non-volatile memory 14 stores programs and data read from an external device 72 via the interface 15, programs and data input via the input device 71, and programs and data acquired from the control device 3 that controls the industrial machine 4 or other devices via the network 5. The programs and data stored in the non-volatile memory 14 may be expanded into the RAM 13 when executed / used. In addition, various system programs such as known analysis programs are written in the ROM 12 in advance.

[0013] The interface 15 is an interface for connecting the CPU 11 of the natural language processing device 1 to an external device 72 such as a USB device. For example, system programs, data, etc. are read from the external device 72. In addition, programs, data, etc. created and edited within the natural language processing device 1 can be stored in external storage means via the external device 72.

[0014] The interface 20 is an interface for connecting the CPU 11 of the natural language processing device 1 to a wired or wireless network 5. The network 5 may be one that communicates using technologies such as serial communication such as RS-485, Ethernet (registered trademark), optical communication, wireless LAN, Wi-Fi (registered trademark), Bluetooth (registered trademark), etc. At least one industrial machine 4 to be controlled, at least one other industrial machine 4, a business application device 2 that responds to inquiries from users, a fog computer 6, a cloud server 7, etc. are connected to the network 5, and data is exchanged between the network 5 and the natural language processing device 1.

[0015] The display device 70 displays various data read into the memory, data obtained as a result of executing programs, etc., output via the interface 17. An input device 71, which is comprised of a keyboard, pointing device, etc., passes instructions, data, etc. based on operations by an operator to the CPU 11 via the interface 18.

[0016] The business application device 2 is an application that uses a text generation AI, an image generation AI, and an LLM to classify and organize existing text knowledge to support the work of users such as operators at manufacturing sites, a chatbot for referring to manuals for industrial machines 4, support for creating NC programs for industrial machines 4 using ChatGPT, support for parameter setting, support for operation, and output some kind of inference result image. The business application device 2 is configured as an application that returns a response sentence to a predetermined query sentence. The business application device 2 includes a generation AI and an LLM that have learned data (text data, program data, image data, etc.) that are responses to the predetermined sentence. These generation AI and LLM can be well-known models such as a Transformer model. For example, the LLM learns the probability that each text will be a response to the predetermined sentence when the predetermined sentence is input. It also further learns the probability that each other text follows the given text. By repeating this learning process, the large-scale language model, when a predetermined sentence is input, outputs a text string with a high probability as a response to the given sentence. In learning, depending on how the document that is a response to a specified sentence is learned, the business application device 2 can be used for purposes such as dialogue, question and answer session, summarizing sentences, editing sentences, translating sentences, converting sentences, modifying sentences, optimizing sentences, interpreting and detecting sentences, recognizing, predicting, judging, generating code, generating images, and making comprehensive judgments.

[0017] The natural language processing device 1 according to the present disclosure is assumed to be connected to a business application device 2 via a network 5. For example, the business application device 2 may have learned the contents of a manual corresponding to a given state of a specific industrial machine 4, for example. Alternatively, the business application device 2 may have learned a command sequence of a machining program for achieving a given purpose, for example, for a sentence indicating a given objective. By narrowing down the information used for learning, it is possible to prepare a business application device 2 that returns responses to inquiries with a certain degree of accuracy.

[0018] 2 is a schematic block diagram showing functions of the natural language processing apparatus 1 according to the first embodiment of the present disclosure. Each function of the natural language processing apparatus 1 according to this embodiment is realized by the CPU 11 included in the natural language processing apparatus 1 shown in FIG. 1 executing a system program and controlling the operation of each part of the natural language processing apparatus 1.

[0019] The natural language processing apparatus 1 of this embodiment includes a parameter input unit 100, a processing data input unit 110, an inference unit 120, an auxiliary information creation unit 130, and an output unit 140. The RAM 13 to the nonvolatile memory 14 of the natural language processing apparatus 1 are also provided with a verification data storage unit 200, which is an area for storing in advance data required to verify processing of a sentence input as processing data, and a business application data storage unit 210, which is an area for storing in advance data related to a business application.

[0020] The parameter input unit 100 accepts input of process setting parameters related to natural language processing from a user. The process setting parameters accepted by the parameter input unit 100 are parameters related to processing to be performed on the target text data accepted by the processing data input unit 110. More specifically, the process setting parameters may be parameters related to a text segmentation method to be executed on the target text data. Furthermore, the process setting parameters may be parameters related to an inference process to be executed based on the target text data. The process setting parameters may directly or indirectly specify these items. For example, the user may directly specify the text segmentation method or the trained model to be used in the inference process. Alternatively, the trained model and text segmentation method used in the business application device 2 may be indirectly specified by selecting the business application device 2 to be used. To enable indirect specification, data related to the trained model and text segmentation method used in each business application may be stored in advance in the business application data storage unit 210. The parameter input unit 100 may accept process setting parameters input via the network 5 from an input device (not shown) included in the control device 3 of the industrial machine 4. In this case, the natural language processing device 1 acts as an intermediary between the operator who operates the control device 3 and the business application device 2. The parameter input unit 100 may also receive the process setting parameters from the input device 71. Furthermore, the parameter input unit 100 may also receive the process setting parameters via a voice input device (not shown). The parameter input unit 100 outputs the received process setting parameters to the inference unit 120 and the auxiliary information creation unit 130.

[0021] The processing data input unit 110 accepts input of processing target text data from a user. The processing target text accepted by the processing data input unit 110 may be natural language text to be processed by the business application device 2. The processing data input unit 110 may accept processing target text data input from an input device (not shown) provided in the control device 3 of the industrial machine 4 via the network 5. In this case, the natural language processing device 1 acts as an intermediary between the operator operating the control device 3 as the user and the business application device 2. The processing data input unit 110 may also accept input of processing target text data via the input device 71. Furthermore, it may also accept input via a voice input device (not shown). The processing data input unit 110 outputs the accepted processing target text data to the inference unit 120 and the auxiliary information creation unit 130.

[0022] The inference unit 120 acquires the result of inference processing based on the text data to be processed input from the parameter input unit 100 and outputs it to the output unit 140. The inference unit 120 according to this embodiment includes a sending unit 122 that sends the text data to be processed to the business application device 2, and a receiving unit 124 that acquires response data from the business application device 2 in response to the text data to be processed. The business application device 2 to which the text data to be processed is to be sent may be received as a processing setting parameter by the parameter input unit 100, for example. The inference unit 120 outputs the received response data to the output unit 140.

[0023] The auxiliary information creation unit 130 creates auxiliary information for verifying the validity of the inference result based on the processing target text data input from the parameter input unit 100 and predetermined verification data stored in the verification data storage unit 200. The auxiliary information creation unit 130, for example, performs text segmentation on the processing target text data using a predetermined text segmentation algorithm. The auxiliary information creation unit 130 also performs text segmentation on the verification target text data stored in the verification data storage unit 200 using a predetermined text segmentation algorithm. When performing text segmentation on the verification target text data, it is desirable to process the data using an algorithm that uses a dictionary that is pre-matched to terms used in the manufacturing site. The text segmentation results of the verification target text data may be created in advance. The auxiliary information creation unit 130 may then create a set of the text segmentation results of the processing target text data, the text segmentation results of the verification target text data, and a comparison result thereof as auxiliary information. The auxiliary information creation unit 130 outputs the created auxiliary information to the output unit 140.

[0024] The output unit 140 outputs the result of inference on the processing target text data input from the inference unit 120 and the auxiliary information created by the auxiliary information creation unit 130. The output unit 140 may transmit the inference result and the auxiliary information via the network 5 to a computer such as the control device 3 that has received the processing target text data. Alternatively, the output unit 140 may output and display the result of inference and the auxiliary information on the display device 70.

[0025] The operation of the natural language processing device will be described below with reference to FIGS. 3 to 7. FIG. 3 is a schematic diagram showing data exchanged between the natural language processing device 1 and other devices. In the example of FIG. 3, an operator operating a control device 3 that controls an industrial machine 4 is assumed as the user of the natural language processing device 1. The user operates an input device (not shown) while looking at a display device (not shown) of the control device 3, and transmits processing setting parameters and processing target text data to the natural language processing device. These processing setting parameters include the specification of a business application device 2. The business application device 2 is assumed to be a general-purpose device such as ChatGPT. The natural language processing device 1 transmits processing target text data input by the user to the business application device 2, receives an inference result, and creates auxiliary information related to the processing target text data. The natural language processing device 1 then outputs the inference result and auxiliary information to the control device 3.

[0026] FIG. 4 is a table diagram showing an example of data related to business applications stored in the business application data storage unit 210. In the example of FIG. 4, data is stored regarding which learned model is used for each business application and which text segmentation method is used. The business application data storage unit 210 may also store, as data, parameters that are set when executing text segmentation processing and inference processing in each business application. These data may be obtained by investigating each business application in advance and stored in the business application data storage unit 210.

[0027] FIG. 5 is a table diagram showing an example of verification target text data stored in the verification data storage unit 200. In the example of FIG. 5, the verification data storage unit 200 also stores text segmentation results obtained by correctly segmenting the verification target text data. In the text segmentation results of FIG. 5, it is assumed that the text is segmented at positions separated by slashes. The verification data storage unit 200 stores, as verification target text data, terms and abbreviations frequently used in manufacturing sites, such as industrial-specific terms or drive control terms, as well as specialized terms used only in the field. When these terms are used in a sentence, performing text segmentation processing using a dictionary containing only general-purpose words may result in erroneous text segmentation results. Such text is pre-stored in the verification data storage unit 200, and correct text segmentation results are also stored as needed.

[0028] FIG. 6 is a screen diagram showing an example of a connection screen of a business application device 2 via a natural language processing apparatus 1. The screen illustrated in FIG. 6 is displayed on the display device of the control device 3. In the example of FIG. 6, business application B is selected as the business application device 2 to be connected to. This selection is used as a processing setting parameter. Business application B is a chat application for Japanese, and as illustrated in FIG. 4, a "Japanese chat dialogue model 1A1" is used as a learning model. Furthermore, a "subword unit algorithm A" is used as a text segmentation method for preprocessing input text to be processed. In this case, as illustrated in FIG. 6, it is assumed that a user makes an inquiry such as "I would like to see the start-up procedure for a new model of machine tool A." This inquiry is transmitted to the natural language processing apparatus 1 as text data to be processed.

[0029] Upon receiving the target text data, the natural language processing apparatus 1 transmits the received target text data to a business application apparatus 2 on which a business application B is running, and acquires an inference result for the target text data. In parallel with this, the auxiliary information creation unit 130 included in the natural language processing apparatus 1 performs text segmentation on the target text data using a subword-based algorithm A, which is a text segmentation method used by the business application B. At this time, it may be possible to determine whether the target text data is included in the target text data, and perform text segmentation processing if it is included. As a result, it is assumed that a text segmentation result such as "I want to see the procedure for starting up a new model of machine tool A" is obtained. The auxiliary information creation unit 130 compares the target text data for verification included in the target text data to determine whether it has been segmented in the same way as the correct text segmentation result for the target text data for verification. In this example, the target text data includes the target text data for verification, "tachiage." If this term is correctly segmented, it is treated as a single term, such as "tachiage." However, as a result of text segmentation of the text data to be processed using the text segmentation method used in business application B, the data is segmented into two terms, i.e., " / tachi / age / ." Therefore, the auxiliary information creation unit 130 adds auxiliary information indicating that there is doubt as to whether the inference was correct, as a result of verifying the validity of the inference result. The created auxiliary information is then output to the output unit 140.

[0030] 7 is a screen diagram showing an example in which the inference result by the business application device 2 and auxiliary information created by the natural language processing device 1 are displayed together. In the example shown in FIG. 7, the auxiliary information related to the processing target text data is displayed in association with the inference result by business application B for the processing target text data. As shown in the example of FIG. 7, if the text segmentation result by business application B for the processing target text data differs from the correct text segmentation result of the verification target text data, auxiliary information is displayed to the effect that there is doubt about the validity of the inference result. Upon seeing such auxiliary information, the user can take action, such as changing the way of making an inquiry or notifying the business application device 2 that there is a problem with the way the text was segmented.

[0031] The natural language processing apparatus 1 according to the present embodiment, which has the above configuration, verifies the validity of the handling of eigenwords and driving relational words in the input text segmentation results, for example, for business applications for industrial fields that use inference processing using LLM, and outputs the verification results to the screen as auxiliary information, allowing the user to confirm whether the output results are based on internal processing that is in line with the user's intention.

[0032] Second Embodiment A natural language processing apparatus according to a second embodiment of the present disclosure will be described below. The natural language processing apparatus 1 according to this embodiment has the same hardware configuration as the natural language processing apparatus 1 according to the first embodiment.

[0033] 8 is a schematic block diagram showing functions of the natural language processing apparatus 1 according to the second embodiment of the present disclosure. Each function of the natural language processing apparatus 1 according to this embodiment is realized by the CPU 11 included in the natural language processing apparatus 1 shown in FIG. 1 executing a system program and controlling the operation of each part of the natural language processing apparatus 1.

[0034] Like the natural language processing apparatus 1 according to the first embodiment, the natural language processing apparatus 1 according to this embodiment includes a parameter input unit 100, a processing data input unit 110, an inference unit 120, an auxiliary information creation unit 130, and an output unit 140. The RAM 13 to the nonvolatile memory 14 of the natural language processing apparatus 1 are provided with a verification data storage unit 200, which is an area for storing in advance data required to verify processing of a sentence input as processing data, and a business application data storage unit 210, which is an area for storing in advance data related to a business application.

[0035] The parameter input unit 100, the processing data input unit 110, and the output unit 140 provided in the natural language processing apparatus 1 according to this embodiment have the same functions as those in the first embodiment.

[0036] The inference unit 120 according to this embodiment has the same functions as the inference unit 120 according to the first embodiment. Furthermore, when the inference result received from the business application device 2 is text data, it outputs the text data of the inference result to the auxiliary information creation unit 130. Furthermore, the auxiliary information creation unit 130 according to this embodiment adds the functions of the auxiliary information creation unit 130 according to the first embodiment, and further performs text segmentation on the text of the inference result input from the inference unit 120 using a text segmentation method specified by the processing setting parameters input from the parameter input unit 100. Then, the text segmentation result is compared with the correct text segmentation result of the verification target text data to verify its validity. The verification result is then added to the auxiliary information, and output to the output unit 140.

[0037] The natural language processing apparatus 1 according to this embodiment performs text segmentation processing on the inference result text data, in addition to the processing target text data input to the business application apparatus 2, when the inference result from the business application apparatus 2 is text data. Then, if the result of segmenting the inference result text data differs from the correct text segmentation result of the verification target text data, auxiliary information is added to indicate that the inference result of the business application apparatus 2 may be invalid. The auxiliary information is then output together with the inference result. This allows the user to confirm whether the output result is based on internal processing in line with their intention.

[0038] [Other Embodiments] In the above-described embodiment, the natural language processing apparatus 1 acts as an intermediary between the terminal operated by the user and another business application apparatus 2. However, for example, the natural language processing apparatus 1 itself may be configured to have functions such as generation AI and LLM in the inference unit 120 and provide functions as a business application.

[0039] Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the individual embodiments described above. Various additions, substitutions, modifications, partial deletions, etc. are possible in these embodiments without departing from the gist of the invention or the idea and intent of the present disclosure derived from the content described in the claims and their equivalents. For example, in the above-described embodiments, the order of each operation and the order of each process are shown as examples and are not limited to these. The same applies when numerical values ​​or mathematical expressions are used in the description of the above-described embodiments.

[0040] The following are supplementary notes related to embodiments of the present disclosure. (Supplementary Note 1) A natural language processing device (1) according to one aspect of the present disclosure includes a verification data storage unit (200) that stores at least verification target text data to be used for processing, a processing data input unit (110) that accepts input of processing target text data, an inference unit (120) that acquires an inference result based on the processing target text data, an auxiliary information creation unit (130) that creates auxiliary information including at least a result of text segmentation of the processing target text data using a predetermined text segmentation algorithm used to obtain the inference result and a result of text segmentation of the verification target text data using the predetermined text segmentation algorithm, and an output unit (140) that outputs the inference result and the auxiliary information.

[0041] (Supplementary Note 2) The auxiliary information created by a natural language processing device (1) according to another aspect of the present disclosure includes a validity determination result obtained by verifying the results of each text segmentation when the processing target text data includes any of the verification target text data. (Supplementary Note 3) The verification target text data stored by a natural language processing device (1) according to another aspect of the present disclosure includes specific terms in an industrial field or drive control terms. (Supplementary Note 4) The natural language processing device (1) according to another aspect of the present disclosure further includes a parameter input unit (100) that accepts input of processing setting parameters, and the inference unit (120) and the auxiliary information creation unit (130) perform processing based on the processing setting parameters.

[0042] (Supplementary Note 5) The processing setting parameters accepted by the natural language processing device (1) according to another aspect of the present disclosure relate to a text segmentation method. (Supplementary Note 6) The text segmentation method used by the natural language processing device (1) according to another aspect of the present disclosure relates to a Japanese text segmentation method, and determines whether segmentation is into words, subwords, or characters.

[0043] (Supplementary Note 7) The inference result by the natural language processing device (1) according to another aspect of the present disclosure includes image data generated by an image generation AI in the processing result. (Supplementary Note 8) The inference result by the natural language processing device (1) according to another aspect of the present disclosure includes text data generated by a text generation AI in the processing result. (Supplementary Note 9) The auxiliary information created by the natural language processing device (1) according to another aspect of the present disclosure includes a text segmentation result of the processing result of the text generation AI.

[0044] (Supplementary Note 10) A computer-readable recording medium according to one aspect of the present disclosure records a program for causing a computer to operate as a verification data storage unit (200) that stores at least verification target text data used in processing, a processing data input unit (110) that accepts input of processing target text data, an inference unit (120) that acquires an inference result based on the processing target text data, an auxiliary information creation unit (130) that creates auxiliary information including at least the result of text segmentation of the processing target text data using a predetermined text segmentation algorithm used to obtain the inference result and the result of text segmentation of the verification target text data using the predetermined text segmentation algorithm, and an output unit (140) that outputs the inference result and the auxiliary information.

[0045] REFERENCE SIGNS LIST 1 Natural language processing device 2 Business application device 3 Control device 4 Industrial machine 5 Network 6 Fog computer 7 Cloud server 11 CPU 12 ROM 13 RAM 14 Non-volatile memory 15, 17, 18, 20 Interface 22 Bus 70 Display device 71 Input device 72 External device 100 Parameter input unit 110 Processing data input unit 120 Inference unit 130 Auxiliary information creation unit 140 Output unit 200 Verification data storage unit 210 Business application data storage unit

Claims

1. A natural language processing device comprising: a verification data storage unit that stores at least verification target text data to be used for processing; a processing data input unit that accepts input of the processing target text data; an inference unit that obtains an inference result based on the processing target text data; an auxiliary information creation unit that creates auxiliary information including at least the result of text segmentation of the processing target text data using a predetermined text segmentation algorithm used to obtain the inference result and the result of text segmentation of the verification target text data using the predetermined text segmentation algorithm; and an output unit that outputs the inference result and the auxiliary information.

2. The natural language processing device according to claim 1, wherein the auxiliary information includes a validity judgment result obtained by verifying the results of each text segmentation when the processing target text data includes any of the verification target text data.

3. The natural language processing device according to claim 1, wherein the text data to be verified includes specific terms in an industrial field or drive control terms.

4. The natural language processing device according to claim 1, further comprising a parameter input unit that accepts input of processing setting parameters, wherein the inference unit and the auxiliary information creation unit perform processing based on the processing setting parameters.

5. The natural language processing apparatus according to claim 4, wherein the processing setting parameters relate to a text segmentation method.

6. The natural language processing apparatus according to claim 5, wherein the text segmentation method is related to a Japanese text segmentation method and determines whether segmentation is to be performed into words, subwords, or characters.

7. The natural language processing device of claim 1, wherein the inference result includes image data generated by an image generation AI in the processing result.

8. The natural language processing device of claim 1, wherein the inference result includes text data generated by a text generation AI in the processing result.

9. The natural language processing device of claim 8, wherein the auxiliary information includes a text segmentation result of the processing result of the text generation AI.

10. A computer-readable recording medium having recorded thereon a program for causing a computer to operate as: a verification data storage unit that stores at least verification target text data used for processing; a processing data input unit that accepts input of processing target text data; an inference unit that obtains an inference result based on the processing target text data; an auxiliary information creation unit that creates auxiliary information including at least the result of text segmentation of the processing target text data using a predetermined text segmentation algorithm used to obtain the inference result and the result of text segmentation of the verification target text data using the predetermined text segmentation algorithm; and an output unit that outputs the inference result and the auxiliary information.

Images