File classification device, file classification method, and file classification program

Abstract

This invention provides a file classification device, a file classification method, and a file classification program that can efficiently classify files. [Solution] The file classification device is characterized by comprising: a file acquisition unit that acquires multiple files of the same type; an integrated file generation unit that generates an integrated file by integrating multiple files; an input unit that inputs classification conditions for the integrated file and multiple files to a generation AI; a result acquisition unit that acquires the results output from the generation AI based on the classification conditions; and a classification unit that classifies files based on the results.

Description

[Technical Field]

[0001] The present invention relates to a file classification device, a file classification method, and a file classification program. [Background technology]

[0002] File classification devices that efficiently classify files such as image files have been known for some time. For example, Patent Document 1 discloses a file classification device that classifies multiple content files, each in which date information related to file creation or editing and classification condition information are recorded, based on the date information and classification condition information.

[0003] In recent years, it has become common practice to digitize (image) the contents of paper documents and other forms using image reading devices such as scanners, and to generate files. The paper documents and other forms to be digitized include text, tables, and drawings. The generated files are usually stored in folders. In this case, it is preferable that each file be grouped (classified) according to its content before it was digitized, i.e., the content of the text, tables, and drawings, and then stored in the folder. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2004-229070 [Overview of the project] [Problems that the invention aims to solve]

[0005] File classification devices are required to efficiently classify files according to their content. This applies not only to image (still image) files, but also to video files, audio files, and other similar files.

[0006] The present invention aims to provide a file classification device, a file classification method, and a file classification program capable of efficiently classifying files. [Means for solving the problem]

[0007] The file classification device according to the present invention is characterized by comprising: a file acquisition unit that acquires multiple files of the same type; an integrated file generation unit that generates an integrated file by integrating multiple files; an input unit that inputs classification conditions for the integrated file and multiple files to a generation AI; a result acquisition unit that acquires the results output from the generation AI based on the classification conditions; and a classification unit that classifies files based on the results.

[0008] Furthermore, in the file classification device according to the present invention, it is preferable that the integrated file generation unit generates an integrated file by degrading each of the multiple files.

[0009] Furthermore, in the file classification device according to the present invention, it is preferable that the integrated file generation unit generates an integrated file that includes identification information corresponding to each of the multiple files.

[0010] Furthermore, in the file classification device according to the present invention, the multiple files are image files, the integrated file generation unit generates an integrated file that includes an image in which each of the images contained in the multiple files is arranged in a tile pattern, and the input unit preferably includes the number of rows and columns of images contained in the integrated file in the classification conditions.

[0011] The file classification method according to the present invention is characterized by acquiring multiple files of the same type, generating a combined file by integrating the multiple files, inputting the combined file and classification conditions for the multiple files into a generating AI, acquiring the results output from the generating AI based on the classification conditions, and classifying the files based on the results.

[0012] The file classification program according to the present invention causes a computer to acquire a plurality of files of the same type, generate an integrated file by integrating the plurality of files, input classification conditions for the integrated file and the plurality of files to a generation AI, acquire the result output from the generation AI based on the classification conditions, and classify the files based on the result.

Effect of the Invention

[0013] According to the present invention, a file classification device, a file classification method, and a file classification program can efficiently classify files.

Brief Description of the Drawings

[0014] [Figure 1] It is a functional block diagram of a file classification system 1 according to Embodiment 1. [Figure 2] It is a flowchart showing the flow of file classification processing. [Figure 3] It is a diagram showing the flow of integrated file generation and an example of an integrated file. [Figure 4] It is a diagram showing an example of classification conditions. [Figure 5] It is a diagram showing an example of the result output from the generation AI. [Figure 6] (a) is a diagram showing an example of integrated file generation according to Modification 1, and (b) is a diagram showing an example of integrated file generation according to Modification 2. [Figure 7] It is a diagram showing the flow of integrated file generation and an example of an integrated file according to Modification 3. [Figure 8] It is a diagram showing the flow of file classification processing according to Modification 4.

Modes for Carrying Out the Invention

[0015] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, it should be noted that the technical scope of the present invention is not limited to the embodiments, and extends to the invention described in the claims and its equivalents.

[0016] (Background of the Invention) When digitizing (imaging) the content described in paper documents such as forms using an image reading device and filing it as electronic data, each file is preferably grouped (classified) according to the content of the text, tables, drawings, etc. before imaging and stored in a folder.

[0017] When files are classified, conventionally, a person checks the files and classifies them as appropriate. However, the number of files stored in a folder may become huge, and there has been a demand to improve the efficiency of the classification work. In contrast, in recent years, the classification of files using generative AI has also been studied. However, even when using generative AI, a person has to describe a prompt in the prompt (instruction text) input field of the generative AI and repeat processes such as uploading one file at a time along with the description of the prompt, which has been costly and time-consuming.

[0018] (Embodiment 1) FIG. 1 is a functional block diagram of a file classification system 1 according to Embodiment 1 of the present invention. The file classification system 1 generates an integrated file by integrating a plurality of files of the same type, inputs classification conditions for the integrated file and the plurality of files to a generative AI, and classifies the files based on the results output from the generative AI.

[0019] The file classification system 1 includes a file classification device 2, a server 3, etc. The file classification device 2 and the server 3 can communicate with each other via a network N.

[0020] The file classification device 2 is an information processing terminal such as a PC (Personal Computer), a mobile phone, a smartphone, a tablet terminal, a game machine, etc. The file classification device 2 may also be a server. The file classification device 2 includes a communication unit 21, a storage unit 22, an operation unit 23, a display unit 24, a processing unit 25, etc. These components are connected via a bus B.

[0021] The communication unit 21 is configured to enable the file classification device 2 to communicate with other devices and includes a communication interface circuit. The communication interface circuit provided by the communication unit 21 is a communication interface circuit such as a wired LAN (Local Area Network) or a wireless LAN. The communication unit 21 receives data from other devices and supplies it to the processing unit 25, and also transmits data supplied from the processing unit 25 to other devices.

[0022] The storage unit 22 is configured for storing programs and data, and includes, for example, semiconductor memory. The storage unit 22 stores programs such as operating system programs, driver programs, and application programs used for processing by the processing unit 25. Programs are installed into the storage unit 22 from computer-readable and non-temporary portable storage media such as CD-ROM (Compact Disc Read Only Memory) and DVD-ROM (Digital Versatile Disc Read Only Memory). Programs may also be installed into the storage unit 22 from an external server via the communication unit 21.

[0023] The operation unit 23 is configured to receive user input to the file classification device 2 and includes, for example, a keyboard, mouse, and keypad. The operation unit 23 may also include a touch panel with a screen such as a liquid crystal display or an organic EL (Electro Luminescence) display, an output interface circuit that outputs image data to the display, and an input interface circuit that acquires signals from the touch panel. In other words, the operation unit 23 may also function as a display unit. The operation unit 23 generates signals corresponding to user input and supplies them to the processing unit 25.

[0024] The display unit 24 is configured for displaying images and includes a display such as a liquid crystal display or an organic EL display. The display unit 24 generates and displays images based on signals supplied from the processing unit 25.

[0025] The processing unit 25 is configured to comprehensively control the operation of the file classification device 2 and comprises one or more processors and their peripheral circuits. For example, the processing unit 25 includes a CPU (Central Processing Unit). The processing unit 25 may also include a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), etc. The processing unit 25 controls the operation of each component and executes various processes so that the various processes of the file classification device 2 are executed in the appropriate procedure based on the program stored in the storage unit 22.

[0026] The processing unit 25 has a file acquisition unit 251, an integrated file generation unit 252, an input unit 253, a result acquisition unit 254, and a classification unit 255 as functional blocks. Each of these units is a functional module realized by the processing unit 25 executing a program. Each of these units may be implemented in the file classification device 2 as firmware.

[0027] Server 3 is an information processing device that provides various services to information processing terminals (client terminals) such as file classification device 2. Server 3 provides services using, for example, generative AI. Generative AI (Artificial Intelligence) is artificial intelligence that learns from large amounts of data and generates new content such as text, images, and music. Generative AI is sometimes also called a Generative Model, Creative AI, Generative Algorithm, AI Creator, AI Generation Engine, or Generative Network. Generative AI is a large-scale language model (LLM) such as "ChatGPT" (registered trademark), "Microsoft Copilot" (registered trademark), "AWS (Amazon Web Services)" (registered trademark), or "Google Cloud" (registered trademark). Generative AI may also be an image generation AI such as "Midjourney," "Stable Diffusion," or "Adobe Firefly" (registered trademark).

[0028] Generative AI generates new content through processes such as data learning, model generation, content generation, and feedback and improvement processes.

[0029] In the data learning process, a generative AI uses a large amount of text and images as "training data" to learn predetermined patterns and rules from this large amount of data. For example, in the case of a generative AI that generates text, it learns grammar, sentence structure, and commonly used word combinations. In the model creation process, the generative AI generates a model that shows the rules and patterns for creating new content based on the learned data. In the content generation process, the generative AI receives prompts from the user and uses the generated model to generate new content based on the input prompts. For example, if part of a sentence is input, the generative AI will generate the rest of the sentence. Also, if a description of an image is input, the generative AI will generate an image corresponding to the input description. In the feedback and improvement process, the generative AI receives evaluation (feedback) on whether the generated content was appropriate or not and improves the model. This enables the generative AI to generate more accurate content. Note that a generative AI only needs to have a content generation process and does not need to have all four processes mentioned above.

[0030] Figure 2 is a flowchart showing the flow of the file classification process performed by the file classification device 2. The file classification process is realized by the processing unit 25 working in cooperation with other components of the file classification device 2 based on a program stored in the storage unit 22.

[0031] First, the file acquisition unit 251 acquires multiple files of the same type (step S101).

[0032] The file retrieval unit 251 retrieves multiple files of the same type from the files stored in the storage unit 22. The file retrieval unit 251 may also retrieve multiple files of the same type from files stored on an external server via the network N and the communication unit 21.

[0033] The files stored in the memory unit 22 include image files, audio files, or text files. Image files include still image files or video files. For example, multiple files of the same type are still image files among the files stored in the memory unit 22. Multiple files of the same type may also be video files or audio files. Multiple files of the same type may also be files with the same image format, such as JPEG (Joint Photographic Experts Group), PNG (Portable Network Graphics), GIF (Graphics Interchange Format), BMP (Bitmap), TIFF (Tagged Image File Format), SVG (Scalable Vector Graphics), EPS (Encapsulated PostScript), PDF (Portable Document Format), or RAW. Multiple files of the same type may also be files with the same file extension. Multiple files of the same type may also be files with the same file format, such as text format or binary format. The following explanation will use the case where multiple files of the same type are still image files as an example.

[0034] Next, the integrated file generation unit 252 generates an integrated file 30 by integrating multiple files (step S102).

[0035] Figure 3 shows the flow of integrated file generation and an example of integrated file 30.

[0036] The integrated file generation unit 252 reads multiple image files 301 and 302 stored in, for example, folder F in the storage unit 22. Image files 301 and 302 are files containing images of paper documents, etc., captured using an image reading device such as a scanner. In the example shown in Figure 3, image file 301 is the image file of image 1, which is a photograph of a document containing text, and image file 302 is the image file of image 2, which is a photograph of a document containing drawings.

[0037] The integrated file generation unit 252 generates an integrated file 30 by integrating multiple files based on the integration conditions. The integration conditions are the conditions for integrating multiple files to generate a single integrated file 30 (or a number of files smaller than the "multiple" in the multiple files). The integration conditions correspond to at least some of the classification conditions described later.

[0038] The integration condition is, for example, to degrade each of the multiple files before integrating them (integration condition 1). That is, the integrated file generation unit 252 generates an integrated file 30 by degrading each of the multiple files. The degradation can be anything that reduces the file size. Degradation can be, for example, compressing the data using a predetermined image compression algorithm including lossless and / or lossy compression, reducing the image resolution by downsampling, or reducing the image color depth (for example, from 24 bits to 16 bits or 8 bits). In the example shown in Figure 3, the integrated file generation unit 252 degrades images 1, 2, etc. by reducing the image resolution.

[0039] The integration condition is, for example, if the files are image files, to integrate them so that the resulting image contains each image from the multiple files arranged in a tiled pattern (integration condition 2). That is, the integrated file generation unit 252 generates an integrated file 30 so that, if the multiple files are image files, the resulting image contains each image from the multiple files arranged in a tiled pattern. A tiled pattern refers to a state in which one or more images are arranged horizontally (rows) and vertically (columns). In the example shown in Figure 3, the integrated file generation unit 252 generates an integrated file 30 so that each image is arranged in n columns vertically and m rows horizontally. The integrated file generation unit 252 may generate multiple integrated files 30 if the number of images exceeds n × m. The combination of the number of images arranged horizontally (rows) and vertically (columns) can be arbitrary; for example, a tiled pattern is an arrangement of 8 columns and 5 rows of images.

[0040] The integration condition is, for example, when integrating images from multiple files so that the resulting image includes a tiled arrangement of each image, it further generates grid lines between each image during the integration (integration condition 3). That is, when the integrated file generation unit 252 generates an integrated file 30 so that the resulting image includes a tiled arrangement of each image from multiple files, it further generates the integrated file 30 so that grid lines are included between each image. The grid lines are lines that clearly indicate that each image is from a different file when the integrated file 30 is generated. The thickness and / or darkness of the grid lines can be any thickness / darkness as long as it clearly indicates that each image is from a different file. However, from the viewpoint of making it easier for the generating AI to understand that each image is from a different file, it is preferable that the thickness / darkness of the grid lines be thicker / darker than the thickness / darkness of the lines contained within each image.

[0041] The integration condition is, for example, to integrate the files so that each of the multiple files contains the corresponding identification information (integration condition 4). That is, the integrated file generation unit 252 generates an integrated file 30 so that each of the multiple files contains the corresponding identification information. The identification information is information for identifying a file. The identification information is, for example, a file name. In the example shown in Figure 3, the file names of each image file (extensions are not shown) are placed below image 1 and image 2 as the identification information 301a for image 1 and the identification information 302a for image 2. Note that the location of the identification information does not have to be below the image; it may be above, to the left, or to the right of the image.

[0042] The integration condition may further include integrating images and their identification information within frames formed by grid lines (integration condition 5). That is, when the integrated file generation unit 252 generates an integrated file 30 such that grid lines are included between each image, it may further generate the integrated file 30 such that images and their identification information are included within frames formed by grid lines.

[0043] The integration condition is, for example, if a file has multiple pages, to integrate the images shown on each page of that file (the entire page) by arranging them in a tile-like manner (integration condition 6). That is, if at least one of the multiple files has multiple pages, the integrated file generation unit 252 generates an integrated file 30 in which the images shown on each page of that file are arranged in a tile-like manner. For example, if an image file has 3 pages, the integrated file generation unit 252 generates an integrated file 30 in which image a1 representing the first page, image a2 representing the second page, and image a3 representing the third page are arranged in a tile-like manner together with the images from the other multiple files.

[0044] The aforementioned integration conditions may be used individually or in any combination. That is, the integrated file generation unit 252 may generate an integrated file 30 by integrating multiple files based on one or more of the integration conditions 1 to 6. In the example shown in Figure 3, the integrated file generation unit 252 generates an integrated file 30 by integrating multiple files based on integration conditions 1, 2, and 4.

[0045] The integration conditions are not limited to the integration conditions 1 to 6 described above. That is, the integrated file generation unit 252 may generate an integrated file 30 by integrating multiple files based on integration conditions other than integration conditions 1 to 6. The integrated file generation unit 252 may also generate an integrated file 30 by integrating multiple files by combining integration conditions other than integration conditions 1 to 6 and at least one of integration conditions 1 to 6.

[0046] Returning to Figure 2, the input unit 253 then inputs the integrated file 30 and classification conditions for multiple files to the generation AI (step S103). When inputting the integrated file 30 to the generation AI, the input unit 253 uploads the integrated file 30 to, for example, the server 3. When inputting the classification conditions to the generation AI, the input unit 253 sends a request signal to the server 3 via the communication unit 21, requesting that the integrated file 30 and classification conditions be input to the generation AI and that the generation AI classify the files. The request signal includes the integrated file 30 and a prompt describing the classification conditions. When the server 3 receives the request signal, it inputs the integrated file 30 and the prompt included in the received request signal to the generation AI.

[0047] Figure 4 shows an example of classification conditions. The classification conditions are those that cause the AI ​​to classify the integrated file 30. At least some of the classification conditions correspond to the integration conditions mentioned above.

[0048] The classification criteria 40 include a description of the contents of the integrated file 30, the rules for classifying the multiple files contained in the integrated file 30, and an example of the classification result output. In the example shown in Figure 4, the description of the contents of the integrated file 30 is written with "#condition", the rules for classifying the files are written with "#group", and the example of the classification result output is written with "#example answer".

[0049] The description of the contents of the integrated file 30 includes the arrangement of each image in the integrated file 30 and the location of the identification information. The description of the contents of the integrated file 30 may also specify that grid lines are included between each image in the integrated file 30. In the example shown in Figure 4, the description of the contents of the integrated file 30, "A maximum of 40 images are arranged in a maximum of 5 rows and 8 columns," corresponds to integration condition 2, and "5 rows and 8 columns" corresponds to the number of rows and columns. Also, the description of the contents of the integrated file 30, "The text below each image is the file name," corresponds to integration condition 4, and "below the image" corresponds to the location of the identification information.

[0050] The rules for classifying files include the groups to be classified (documents, drawings, tables, etc.) and the characteristics of each group. A "Other" category, which does not belong to any group, may also be defined. This helps prevent files with unexpected content from being incorrectly classified into the wrong group.

[0051] The example of the classification result output shows an example of the information output by the generating AI. An example of the classification result output is a list containing the names of each group and the identification information (file names) of the files belonging to each group. This allows the file classification device 2 to have the generating AI output results in a consistent manner. The classification condition 40 may include instructions such as not including any explanation other than the result in the classification result. This makes it easier for the file classification device 2 to classify files based on the results output by the generating AI.

[0052] The description of the contents of the integrated file 30 may specify that grid lines are included between each image contained in the integrated file 30.

[0053] Returning to Figure 2, the result acquisition unit 254 then acquires the results output from the generating AI based on the classification conditions (step S104). The server 3 acquires the classification results from the generating AI and transmits the result information, including the acquired classification results, to the file classification device 2. The result acquisition unit 254 receives the result information from the server 3 via the communication unit 21 and acquires the classification results included in the result information.

[0054] Figure 5 shows an example of the output from the generating AI. The output from the generating AI is based on the output example of the classification result defined in the classification conditions. In the example shown in Figure 5, the identification information of each file is classified into "document" and "drawing," respectively.

[0055] Returning to Figure 2, the classification unit 255 then classifies the files based on the results output from the generating AI (step S105). The classification unit 255 classifies the files based on the identification information contained in the results 50 output from the generating AI. The classification unit 255 then distributes and stores each file in folders (for example, a folder for document files or a folder for drawing files) that have been pre-classified within the file classification device 2, based on the file names of each file contained in the results 50. This completes the file classification process.

[0056] As explained above, the file classification device 2 generates a combined file 30 by integrating multiple files. The file classification device 2 also inputs the combined file 30 and classification conditions 40 for the multiple files into the generating AI. The file classification device 2 then classifies the files based on the results 50 output from the generating AI based on the classification conditions 40. This allows the file classification device 2 to classify multiple files at once with a single prompt (instruction) when using the generating AI. Therefore, the file classification device 2 can classify files efficiently.

[0057] The generation AI may have a daily usage limit (number of commands, image size, etc.) when using the service, and uploading files one by one can quickly reach this limit. In contrast, File Classifier 2 enables the classification of multiple files with a single prompt. Furthermore, File Classifier 2 can obtain results equivalent to multiple files in a single response (output) from the generation AI. This also contributes to improving the speed of classification processing in File Classifier 2. As a result, File Classifier 2 can significantly improve the efficiency of file classification work.

[0058] Preferably, the integrated file generation unit 252 generates an integrated file 30 by degrading each of the multiple files. Each file included in the integrated file 30 may be of low quality, as long as it can be classified by the generation AI. By degrading and integrating each file, the integrated file generation unit 252 can reduce the size of the integrated file 30. As a result, the file classification device 2 can efficiently classify files even when a daily usage limit is set when using the generation AI service, as described above.

[0059] Preferably, the integrated file generation unit 252 generates an integrated file 30 that includes identification information 30a corresponding to each of the multiple files. Generally, the clearer the prompt, the more the generating AI can output results that are in line with the prompt. Since the identification information 30a corresponds to each of the multiple files contained in the integrated file 30, the generating AI can classify the files with high accuracy using the identification information 30a. As a result, the file classification device 2 can classify files with higher accuracy.

[0060] Preferably, the integrated file generation unit 252 generates an integrated file 30 that includes an image in which each image from the multiple files is arranged in a tile-like pattern. The input unit 253 includes the matrix of images contained in the integrated file 30 in the classification condition 40. This allows the file classification device 2 to efficiently arrange multiple image files in the integrated file 30, thereby increasing the number of image files that can be processed together. Furthermore, because multiple images are arranged regularly in the integrated file 30, the generating AI can classify each file with high accuracy. As a result, the file classification device 2 can classify files more efficiently and with higher accuracy.

[0061] Preferably, the integrated file generation unit 252 generates an integrated file 30 based on integration conditions that correspond to at least a portion of the classification conditions 40. If the integration conditions used when generating the integrated file 30 from multiple files differ from the classification conditions used when having the generation AI classify the files, the generation AI may not be able to accurately classify each file included in the integrated file 30. The file classification device 2 enables the generation AI to accurately classify the files included in the integrated file 30 because the integration conditions correspond to at least a portion of the classification conditions. As a result, the file classification device 2 can classify files with higher accuracy.

[0062] Preferably, when the integrated file generation unit 252 generates an integrated file 30 that includes an image in which each image from multiple files is arranged in a tile pattern, it further generates the integrated file 30 so that grid lines are included between each image. The generation AI can classify files with high accuracy using the grid lines in the integrated file 30. As a result, the file classification device 2 can classify files with higher accuracy.

[0063] Preferably, when the integrated file generation unit 252 generates an integrated file 30 that includes an image in which each image contained in multiple files is arranged in a tile pattern, and when the integrated file 30 is generated so that grid lines are included between each image, it further generates the integrated file 30 so that the image and the identification information of the image are included inside the frame formed by the grid lines. Because the identification information is included within the grid lines, the generating AI can classify the files with high accuracy using the grid lines and the identification information. As a result, the file classification device 2 can classify files with higher accuracy.

[0064] Preferably, if the file has multiple pages, the integrated file generation unit 252 generates an integrated file 30 in which the images shown on each page of the file are arranged in a tiled pattern. This allows the file classification device 2 to efficiently classify files, including the images corresponding to each page of the file having multiple pages.

[0065] (Modification 1 and Modification 2 of Embodiment 1) In Embodiment 1, the multiple files were described as mainly being image files. However, the multiple files may also be, for example, video files or audio files.

[0066] Figure 6(a) shows an example of integrated file generation according to Modification 1. In Modification 1, the file classification device 2 classifies video files as multiple files of the same type. The file classification device 2 may also classify video files that are mutually identical in video format, such as MPEG (Moving Picture Experts Group), WMV (Windows Media Video), and rv (Real Video), as multiple files of the same type.

[0067] The file acquisition unit 251 acquires multiple video files of the same file type from the files stored in the storage unit 22. The file acquisition unit 251 may also acquire multiple video files of the same file type from files stored on an external server via the network N and the communication unit 21.

[0068] The integrated file generation unit 252 reads multiple video files (video data) 311 and 312 stored in, for example, folder F in the storage unit 22. Video file 311 is, for example, a video file recording a forest scene, and video file 312 is, for example, a video file recording a sea scene.

[0069] The integrated file generation unit 252 generates an integrated file 31 by integrating multiple video files 311 and 312. The integrated file generation unit 252 generates the integrated file 31 by combining the multiple video files 311 and 312 so that they can be played consecutively. The integration condition when generating the integrated file 31 is, for example, to unify the video duration of each video file (video data) to the video duration of the shortest video file among the multiple video files. Unifying the video duration is, for example, by editing the video of each video file to a predetermined length (e.g., 10 minutes). Editing is, for example, changing the playback speed of the video duration, deleting a part of the video, etc. The integration condition may also be to insert a specific image indicating a delimiter between each video file. The specific image is, for example, an image showing only one specific color such as blue or red. The specific image may also be an image showing a symbol such as "circle (〇)" or "cross (×)". The specific image may also be an image showing identification information (e.g., file name). The integration condition may also be to insert a specific video indicating a delimiter between each video file. A specific video is, for example, a video of a few seconds (e.g., 1 second) showing a specific color as mentioned above. A specific video may also be a video of a few seconds showing the symbol mentioned above. A specific video may also be a video of a few seconds showing identifying information.

[0070] The integrated file generation unit 252 degrades and integrates multiple video files by compressing the data, reducing the image resolution, or decreasing the image color depth, similar to how it integrates still image files.

[0071] When the integrated file generation unit 252 generates an integrated file 31 by integrating multiple video files 311, 312, it may generate the integrated file 31 by converting each of the multiple video files into an image file. Converting video files to image files can be done, for example, by sampling the video file at predetermined time intervals and extracting still images. The integrated file generation unit 252 generates an image by arranging the still images from each sampling point, for example, in a tile pattern.

[0072] The input unit 253 inputs classification conditions for the integrated file 31 and multiple video files to the generating AI in the same manner as in step S103 described in Embodiment 1.

[0073] The classification conditions, similar to those described in Embodiment 1, include a description of the contents of the integrated file 31, rules for classifying the multiple files contained in the integrated file 31, and an example of the output classification result. The description of the contents of the integrated file 31 indicates, for example, that the integrated file 31 is a video file, that the video duration of each video file is a predetermined duration (or that specific images are included between each video file), etc. The rules for classifying the multiple files contained in the integrated file 31 indicate, for example, that the files are classified by the object being imaged, and the characteristics of the color and movement of each object, etc. The example of the output classification result is the same as the classification conditions described in Embodiment 1.

[0074] Figure 6(b) shows an example of integrated file generation according to Modification 2. In Modification 2, the file classification device 2 classifies audio files as multiple files of the same type. The file classification device 2 may also classify audio files that are mutually identical in format, such as MP3 (MPEG-1 Audio Level 3), WMA (Windows Media Audio), and ra (Real Audio), as multiple files of the same type.

[0075] The file acquisition unit 251 acquires multiple audio files of the same type from the files stored in the storage unit 22. The file acquisition unit 251 may also acquire multiple audio files of the same type from files stored on an external server via the network N and the communication unit 21.

[0076] The integrated file generation unit 252 reads multiple audio files (audio data) 321 and 322 stored in, for example, folder F in the storage unit 22. Audio file 321 is, for example, an audio file containing a recording of the sound of a road during the day, and audio file 322 is, for example, an audio file containing a recording of the sound of a road at night.

[0077] The integrated file generation unit 252 generates an integrated file 32 by integrating multiple audio files 321 and 322. The integrated file generation unit 252 generates the integrated file 32 by combining the multiple audio files 321 and 322 so that they can be played consecutively. The integration condition when generating the integrated file 32 is, for example, to unify the audio time in each audio file (audio data) to the audio time of the audio file with the shortest audio time among the multiple audio files. Unifying the audio time is, for example, by editing the audio of each audio file to a predetermined length. Editing is, for example, changing the playback speed of the audio time, deleting a part of the audio, etc. The integration condition may also be to insert a specific sound to indicate a delimiter between each audio file. The specific sound may be, for example, an inaudible sound (ultrasonic sound). The specific sound may also be an audio that reads out identification information (for example, a file name).

[0078] The integrated file generation unit 252 degrades the quality of multiple audio files by compressing the data, similar to how it does when integrating still image files or video files, before integrating them.

[0079] The input unit 253 inputs the classification conditions for the integrated file 32 and the multiple audio files to the generating AI in the same manner as in step S103 described in Embodiment 1.

[0080] The classification conditions, similar to those described in Embodiment 1, include a description of the contents of the integrated file 32, rules for classifying the multiple files contained in the integrated file 32, and an example of the output classification result. The description of the contents of the integrated file 32 indicates, for example, that the integrated file 32 is an audio file, that the audio duration of each audio file is a predetermined duration (or that a specific sound is included between each audio file), etc. The rules for classifying the multiple files contained in the integrated file 32 indicate, for example, that the files are classified by the target sound recorded, and the frequency and amplitude characteristics of each target sound, etc. The example of the output classification result is the same as the classification conditions described in Embodiment 1.

[0081] (Modification 3 of Embodiment 1) In Embodiment 1, the integrated file generation unit 252 generated an integrated file 30 that included identification information corresponding to each of the multiple files. However, the identification information does not necessarily have to be included in the integrated file 30.

[0082] Figure 7 shows the flow of integrated file generation and an example of an integrated file related to the modified example 3. The integrated file generation unit 252 generates an integrated file 33 based on integration conditions other than the aforementioned integration condition 4. In other words, the integrated file 33 does not contain identification information. In the example shown in Figure 7, the multiple files are image files, and the integrated file generation unit 252 generates the integrated file 33 based on at least integration condition 2.

[0083] The integrated file generation unit 252 generates an identification list L containing identification information corresponding to each of the multiple files when generating the integrated file 33. Preferably, the identification list L is a list in which the identification information is arranged in the order in which the multiple files were integrated. That is, when generating the integrated file 33, the integrated file generation unit 252 generates an identification list L in which the identification information corresponding to each of the multiple files is arranged in the order in which the multiple files were integrated. For example, if, based on integration condition 2, the images of the multiple files are arranged sequentially from left to right in Figure 7, starting from the first row and first column of a tile, the integrated file generation unit 252 generates an identification list L containing the m rows and n columns of the tile. i row n i The identification information of the image in column {(m i -1) × n + n i An identification list L is generated, arranged as number}. For example, if the identification information is the file name, and the image of the file named "10-636-03" is a tiled image in the first row and first column, then "10-636-03" will be listed as the first item in the identification list L.

[0084] The input unit 253 inputs classification conditions including the integrated file 33 and the integrated order list into the generation AI. The integrated order list is an example of an output example of the classification result, and is a list that associates the name of each group with the integration order of the files classified into each group. For example, when the integrated file 33 is generated based on the integration condition 2, the integrated order list is a list that associates the classified groups with the mth row and nth column in the tile shape of m rows and n columns. i rows n i column image as {(m i -1)×n + n i}th and {(m i -1)×n + n i}th. In this case, the integrated order list describes, for example, "Document No. 1", "Drawing No. 2", etc. (or "No. 1 Document", "No. 2 Drawing", etc.).

[0085] The result acquisition unit 254 acquires the result (output result) output from the generation AI based on the classification conditions including the integrated order list. Since the output result is output from the generation AI based on the integrated order list, which is an example of an output example of the classification result, the output result includes the integrated order list.

[0086] The classification unit 255 classifies files based on the integrated order list and the identification list L included in the output result output from the generation AI. For example, the classification unit 255 associates the integrated order list (e.g., "Document No. 1") with the identification list L (e.g., "10-636-03", which is the first file in the identification list L) to classify the files.

[0087] In the file classification device according to Modification 3, the integrated file generation unit 252 generates an identification list L containing identification information corresponding to each of the multiple files when generating the integrated file 33. Preferably, the integrated file generation unit 252 generates an identification list L in which the identification information corresponding to each of the multiple files is arranged in the order in which the multiple files were integrated. The input unit 253 inputs the classification conditions, which include the integrated file 33 and the integration order list, to the generation AI. The result acquisition unit 254 acquires the output results output based on the classification conditions that include the integration order list. The classification unit 255 classifies the files based on the integration order list and the identification list L included in the output results. As a result, this file classification device can classify files with high accuracy even without including identification information in the integrated file 33. This is particularly effective, for example, when it is not easy to include identification information in the integrated file 33.

[0088] Furthermore, the file classification process described in Modification Example 3 above can also be applied when multiple files are video files or audio files. In this case, the identification list L can also be described as a list in which the identification information corresponding to each of the multiple files is arranged in chronological order when the multiple files were merged. The merged order list can also be described as a list that associates the names of each group with the chronological order in which the files classified into each group were merged.

[0089] (Further modification of Modification 3 of Embodiment 1) If the multiple files are video files or audio files, the input unit 253 may input classification conditions to the generating AI that include time information indicating the playback time (video time or audio time) of the integrated file 33 and the time (time) when each of the multiple files is integrated, as a description of the contents of the integrated file 33.

[0090] For example, if video file A (or audio file D) has a playback time of 3 minutes, video file B (or audio file D) has a playback time of 1 minute, and video file C (or audio file F) has a playback time of 2 minutes, and these video files or audio files are combined in this order to generate a combined file 33, then the time information will include that the playback time of the combined file 33 is 6 minutes, that video files A, B, and C (or audio files D, E, and F) are combined in this order, and that each video in each video file (or each audio in each audio file) starts at playback start time 0 minutes, 3 minutes, and 4 minutes. In this case, the output from the generating AI will include time classification results in which groups are classified in relation to the playback start time (for example, start times 0 to 3 minutes are "Generating AI", start times 3 to 4 minutes are "GIS (Geographic Information System)", start times 4 to 6 minutes are "Measurement", etc.). In other words, the result acquisition unit 254 acquires results that include time classification results output from the generating AI, in which groups are classified in relation to the playback start time.

[0091] The classification unit 255 classifies (stores) video files A, B, and C (or audio files D, E, and F) into their respective folders (for example, "Generation AI Folder", "GIS Folder", "Measurement Folder", etc.) based on the results, including the time classification results.

[0092] In the modified file classification device, the input unit 253 inputs classification conditions to the generating AI that include time information indicating the playback time of the integrated file 33 and the combined time of each file. The result acquisition unit 254 acquires the results output from the generating AI, which include time classification results in which groups are classified in relation to the playback start time. The classification unit 255 classifies the files based on the results including the time classification results. As a result, this file classification device can classify files with high accuracy even for files such as video files and audio files, for which it is not easy to include identification information in the integrated file 33.

[0093] (Modification 4 of Embodiment 1) Figure 8 shows the flow of the file classification process according to Modification 4. The file classification device according to Modification 4 classifies (stores) files that have multiple pages and in which each page (each image) is classified into a different group (hereinafter referred to as a "mixed file") into a folder that stores groups containing characteristics of multiple groups (hereinafter referred to as a "mixed folder").

[0094] The integrated file generation unit 252 integrates multiple pages Mpf1 and Mpf2 contained in a file having multiple pages (hereinafter referred to as a "multi-page file") Mpf, together with other files to generate an integrated file 34. The integrated file generation unit 252 generates the integrated file 34 based on, for example, integration condition 6. The integrated file generation unit 252 may further generate the integrated file 34 based on at least one of integration conditions 1 to 5.

[0095] When an integrated file 34 is generated based on integration condition 4, the integrated file generation unit 252 generates the integrated file 34 by including the page number corresponding to each page Mpf1, Mpf2 of the multi-page file Mpf in the identification information of each page (each image). For example, if the identification information is the file name, the integrated file generation unit 252 generates the file name by adding the page number before or after the file name. For example, if the file name of the multi-page file Mpf is "10-636-03" and one of the pages Mpf1, Mpf2 is page 1, the identification information of page Mpf1 will be indicated as "page 1_10-636-03" or "10-636-03_page 1", etc.

[0096] If the integrated file 34 is generated based on integration condition 4, the input unit 253 inputs the integrated file 34 and the classification conditions for the multiple files to the generation AI in the same manner as in step S103 described in Embodiment 1. The result acquisition unit 254 acquires the results output from the generation AI based on the classification conditions in the same manner as in step S104 described in Embodiment 1. The results output from the generation AI include identification information that contains the page numbers corresponding to each page Mpf1, Mpf2 of the multi-page file Mpf, and these are classified into groups.

[0097] If the integrated file 34 is generated based on an integration condition other than integration condition 4, the integrated file generation unit 252 generates an identification list L in which the page number corresponding to each page of the multi-page file Mpf is included in the identification information of each page when generating the integrated file 34. For example, if the multi-page file Mpf is the first to be integrated into the integrated file 34, and the identification information is the file name, and the file name of the multi-page file Mpf is "10-636-03", then the first entry in the identification list L will be "10-636-03 Page 1", the second entry in the identification list L will be "10-636-03 Page 2", and so on.

[0098] If an identification list L has been generated, the input unit 253 inputs the classification conditions, which include the integrated file 34 and the integrated order list, to the generating AI in the same manner as in Modification 3 of Embodiment 1. The result acquisition unit 254 acquires the results output from the generating AI based on the classification conditions, which include the integrated order list, in the same manner as in Modification 3 of Embodiment 1.

[0099] The classification unit 255 stores multiple files, including the multi-page file Mpf, into folders, including the mixed folder MFo, based on the results output from the generating AI and identification information containing the page number corresponding to each page, or an identification list L that includes the page number corresponding to each page in the identification information of each page.

[0100] If, in the output from the generating AI, each page of the multi-page file Mpf is classified into a different group, the classification unit 255 stores the multi-page file Mpf in the mixed folder MFo. If, in the output from the generating AI, each page of the multi-page file Mpf is not classified into a different group, that is, if each page of the multi-page file Mpf is classified into the same group, the classification unit 255 stores the multi-page file Mpf in the folder corresponding to that group (for example, "Documents folder" or "Drawings folder").

[0101] The classification unit 255 may generate split files DFi1 and DFi2 by splitting each page Mpf1 and Mpf2 of the multi-page file Mpf stored in the mixed folder MFo. The classification unit 255 may further store the split files DFi1 and DFi2 in folders corresponding to the group of each page, based on the results output from the generation AI, which include identification information containing the page number corresponding to each page, or an identification list L that includes the page number corresponding to each page in the identification information of each page. For example, if, in the results output from the generation AI, page Mpf1 corresponding to split file DFi1 is classified into the "Drawing" group, while page Mpf2 corresponding to split file DFi2 is classified into the "Document" group, the classification unit 255 will store split file DFi1 in the "Drawing" folder and split file DFi2 in the "Document" folder.

[0102] In the file classification device according to Modification 4, the integrated file generation unit 252 generates an integrated file 34 by including the page numbers corresponding to each page Mpf1, Mpf2 of the multi-page file Mpf in the identification information of each page, or, when generating the integrated file 34, generates an identification list L that includes the page numbers corresponding to each page of the multi-page file Mpf in the identification information of each page. The classification unit 255 then classifies multiple files, including the multi-page file Mpf, based on the results output from the generation AI and the identification information containing the page numbers corresponding to each page, or the identification list L that includes the page numbers corresponding to each page in the identification information of each page. As a result, this file classification device can classify files with high accuracy even when a file has multiple pages and each page of the file is classified into a different group.

[0103] Preferably, if each page of the multi-page file Mpf is classified into a different group, the classification unit 255 stores the multi-page file Mpf in a mixed folder MFo. More preferably, the classification unit 255 splits each page Mpf1 and Mpf2 of the multi-page file Mpf stored in the mixed folder MFo, and stores the split files DFi1 and DFi2 in folders corresponding to each page group, based on the results output from the generation AI and identification information containing the page number corresponding to each page, or an identification list L that includes the page number corresponding to each page in the identification information of each page. This enables the file classification device to classify files with high accuracy even when each page of a multi-page file is classified into a different group.

[0104] (Other modifications of Embodiment 1) The following variations may be applied to file classification system 1.

[0105] In the above explanation, it was assumed that server 3 has a generation AI, but the example is not limited to this. For example, file classification device 2 may also have a generation AI. In this case, in step S103 above, the input unit 253 inputs the integrated file and prompt to the generation AI. Also, in step S104 above, the result acquisition unit 254 acquires the classification result from the generation AI. This makes it possible for file classification device 2 to efficiently classify files in a so-called standalone manner.

[0106] In the above explanation, when multiple files are image files, the integrated file generation unit 252 generates an integrated file that includes an image in which each image from the multiple files is arranged in a tiled pattern. However, the system is not limited to this example. For example, the integrated file generation unit 252 may generate an integrated file that includes an image in which each image from the multiple files is arranged in a staggered pattern or randomly. In this case as well, the input unit 253 inputs the arrangement of each image in the integrated file (for example, "The images are arranged in a staggered pattern," or "The images are arranged randomly," etc.) as a classification condition to the generation AI, enabling the generation AI to classify the integrated file. In other words, the file classification device 2 can classify files efficiently.

[0107] Furthermore, the integrated file generation unit 252 may generate an integrated file containing multiple pages. In this case, the integrated file generation unit 252 generates the integrated file such that each page of the integrated file contains an image from each file to be classified. In this case, the classification condition is defined as the inclusion of an image to be classified on each page of the integrated file.

[0108] Furthermore, the integrated file generation unit 252 may generate an integrated file by integrating multiple thumbnail images (reduced versions of the original images) generated by processes other than those performed by the integrated file generation unit 252. For example, if multiple image files in a folder are displayed as thumbnail images in a tiled pattern, the integrated file generation unit 252 generates an integrated file by taking a screen capture of the tiled thumbnail images. In this case, the integrated file generation unit 252 can generate an integrated file using the arrangement of the tiled thumbnail images. This reduces the processing burden on the file classification device for generating integrated files.

[0109] In the explanation above, it was assumed that the integrated file is not included in the prompts input to the generating AI, but the integrated file may be included in the prompts input to the generating AI.

[0110] In the above description, the classification unit 255 classified the files based on the results output from the generating AI. The classification unit 255 may further add the name of the group to be classified to the file's identification information (e.g., file name) when classifying files. For example, if the group name is added to the file name of each file in a set of multiple files, the classification unit 255 adds the group name before or after the file name. In this case, the file name would be, for example, "Document_10-636-03" or "10-636-03_Document" if the file named "10-636-03" is a document file. If there are split files generated by splitting a multi-page file, as described in Modification 4 of Embodiment 1, the classification unit 255 may add the name of the group to be classified and the page number to the identification information of the split file.

[0111] Furthermore, the classification unit 255 may generate a correspondence table that associates the group to be classified with identification information (e.g., file name). The classification unit 255 generates the correspondence table such that, for example, the first column of the table is the group and the second column is the file name. In this case, if the file named "10-636-03" is a document file, then the correspondence table will contain "Document" in the first row, first column, and "10-636-03" in the second row, second column. If there are split files generated by splitting a multi-page file, as described in Modification 4 of Embodiment 1, the classification unit 255 may generate a correspondence table that associates the name, identification information, and page number of the group to be classified. In this case, the page number is the page number when the split file was included in the multi-page file.

[0112] Those skilled in the art will understand that various changes, substitutions, and modifications can be made without departing the scope of the present invention. For example, the embodiments and modifications described above may be combined as appropriate within the scope of the present invention. [Explanation of Symbols]

[0113] 1. File Classification System 2 File Classification Device 25 Processing Unit 251 File Acquisition Section 252 Integrated File Generation Unit 253 Input section 254 Result acquisition part 255 Classification Department 3. Server (Generating AI)

Claims

1. A file retrieval unit that retrieves multiple files of the same type, A combined file generation unit that generates a combined file by integrating the aforementioned multiple files, An input unit that inputs the classification conditions for the integrated file and the multiple files into the generating AI, A result acquisition unit that acquires the results output from the generating AI based on the classification conditions, A classification unit that classifies the files based on the above results, A file classification device characterized by comprising the following features.

2. The integrated file generation unit generates the integrated file by degrading each of the plurality of files. The file classification device according to claim 1.

3. The integrated file generation unit generates the integrated file such that it contains identification information corresponding to each of the multiple files. The file classification device according to claim 1 or 2.

4. The aforementioned files are image files, The integrated file generation unit generates the integrated file such that it includes an image in which each of the images contained in the plurality of files is arranged in a tile pattern. The input unit includes the number of rows and columns of images included in the integrated file in the classification conditions. The file classification device according to claim 1 or 2.

5. Retrieve multiple files of the same type, A combined file is generated by integrating the aforementioned multiple files. The classification conditions for the integrated file and the multiple files are input to the generating AI. Based on the classification conditions, the results output from the generating AI are obtained. Based on the above results, classify the files. A file classification method characterized by the following features.

6. On the computer, Retrieve multiple files of the same type, A combined file is generated by integrating the aforementioned multiple files. The AI ​​is given the classification conditions for the integrated file and the multiple files to input. Based on the classification conditions, the results output from the generating AI are obtained. Based on the above results, classify the files. A file classification program characterized by the following features.

Images