Information processing apparatus, information processing method, and program

Abstract

To improve retrieval of document image files.SOLUTION: An information processing apparatus according to the present invention has classifying means for classifying a document type of a document image. The information processing apparatus also has extracting means for extracting a character string group corresponding to item information from the document image. The information processing apparatus also has storing means for storing the document image with giving a file name to the document image. The classifying means classifies a document image, which corresponds to a plurality of document types (e.g., a multi-purpose form), into multiple document types. When the document image is classified into multiple document types, the storing means assigns a plurality of different file names to the same document image using the character string group extracted by the extracting means, and stores the document image.SELECTED DRAWING: Figure 7

Description

[Technical field]

[0001] The present invention relates to techniques for extracting information from document images. [Background technology]

[0002] Conventionally, there is a technique for determining the document type (e.g., invoice, quotation, delivery note) of an input document image, and a technique for extracting character strings corresponding to predetermined item information (e.g., title, document number, issue date, company name, amount) written in the document image. Patent Document 1 discloses a technique for determining the document type of a document image, comparing the similarity with characteristic information of a folder for storing the document image, and setting a link to the file of the document image for a folder whose similarity is equal to or greater than a threshold. According to Patent Document 1, for example, for a document image file of a document that combines multiple document types (hereinafter referred to as a combined form), such as a "delivery note and invoice," a link can be set to both the "delivery note" folder and the "invoice" folder. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2004-220347 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology of Patent Document 1, when a file is classified and stored in a plurality of folders whose similarity to the file is equal to or greater than a threshold, a search must be performed on the group of files classified into folders having characteristic information, which reduces the searchability of the file. Specifically, for example, when a "delivery note / invoice" is stored in a specified folder and then the file is referred to as a "delivery note", it takes time and effort to find the file from among all the files in the "delivery note" folder, which reduces the searchability. Similarly, for example, when a "delivery note / invoice" is stored in a specified folder and then the file is referred to as an "invoice", it takes time and effort to find the file from among all the files in the "invoice" folder, which reduces the searchability. In addition, in the technology of Patent Document 1, it is possible to classify and store files by subdividing the characteristic information of the folder, but the more the characteristic information is subdivided, the more complex the folder structure becomes, which requires time and effort to refer to multiple folders, which reduces the searchability.

[0005] SUMMARY OF THE PRESENT EMBODIMENT An object of the present invention is to improve the searchability of document image files. [Means for solving the problem]

[0006] The information processing device of the present invention has a classification means for classifying the document type of a document image, an extraction means for extracting a group of character strings corresponding to item information from the document image, and a storage means for assigning a file name to the document image and storing it, and is characterized in that when the document image is classified into a plurality of document types by the classification means, the storage means assigns a plurality of different file names to the document image using the group of character strings extracted by the extraction means. Effect of the Invention

[0007] According to the present invention, it is possible to improve the searchability of document image files. [Brief description of the drawings]

[0008] [Figure 1]FIG. 1 is a diagram illustrating an example of an overall configuration of an information processing system. [Diagram 2] FIG. 2 is a diagram illustrating an example of the configuration of each device. [Diagram 3] FIG. 1 is a diagram showing a usage sequence of an information processing system. [Figure 4] 1 is a flowchart for generating a machine learning model. [Diagram 5] 11 is a flowchart for storing a link to a file of a document image. [Figure 6] 11 is a flow chart for storing a file copy of a document image. [Figure 7] 11 is a flowchart for adding metadata to a document image file. [Figure 8] FIG. 13 is a diagram showing an example of an image of a document (multipurpose form) that covers a plurality of document types. [Figure 9] FIG. 11 is a diagram showing an example of a document type determination result. [Figure 10] FIG. 1 is a diagram illustrating an example of input and output data of a machine learning model. [Figure 11] FIG. 10 is a diagram illustrating a method for storing a document image file. [Figure 12] 11 is an explanatory diagram of the relationship between a character string included in a title of a document image and a document type. [Figure 13] FIG. 13 is a diagram showing an example of an image of a document (multipurpose form) that covers a plurality of document types. [Figure 14] FIG. 10 is a diagram illustrating a method for storing a document image file. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0010] [Embodiment 1] <Information Processing System> FIG. 1 is a diagram showing an example of the overall configuration of an information processing system. As shown in FIG. 1, the information processing system 100 is composed of an information processing device 101, a learning device 102, and an information processing server 103. The information processing device 101, the learning device 102, and the information processing server 103 are connected to each other via a network 104. The information processing system 100 may be composed of one each of the information processing device 101, the learning device 102, and the information processing server 103, or may be composed of multiple each of them. For example, the information processing server 103 may be composed of a first server device having high-speed calculation resources and a second server device having large-capacity storage, and may be configured to be connected to each other via the network 104.

[0011] The information processing device 101 is realized by an MFP (Multi-Function Peripheral) having multiple functions such as printing, scanning, and faxing, and functions as an image acquisition unit 151. The information processing device 101 optically reads an original 111 printed on a storage medium such as paper using the image acquisition unit 151, and performs a predetermined scan process to generate a document image 113 and transmit it to the information processing server 103.

[0012] Furthermore, the information processing device 101 receives FAX data 112 transmitted from a FAX transmitter (not shown) by the image acquisition unit 151, generates a document image 113 by performing a predetermined FAX process, and transmits the document image 113 to the information processing server 103. The information processing device 101 may be realized by a PC (Personal Computer) or the like other than an MFP having the above-mentioned scan and FAX functions. Specifically, the document image 113, such as PDF or JPEG, generated by using a document creation application operating on the PC serving as the information processing device 101 may be transmitted to the information processing server 103.

[0013] The learning device 102 has functions as a generation unit 152 and a learning unit 153. The generation unit 152 generates learning data by assigning a correct answer label (first correct answer label) of the document type of the document image to a plurality of document image samples 114 and assigning a correct answer label (second correct answer label) to a character string to be extracted from a group of character strings included in the document image. In addition, the learning unit 153 generates a machine learning model 115 for determining the document type of the document image and extracting a group of character strings to be extracted from a group of character strings included in the document image by having a machine learning model learn the learning data generated by the generation unit 152.

[0014] In this embodiment, the learning device 102 acquires document image samples 114 of different types of forms, such as invoices, estimates, purchase orders, and delivery notes, and generates a document type determination model, which is a machine learning model 115 for estimating the document types to which these are classified. The learning device 102 also generates an item information extraction model, which is a machine learning model 115 for extracting character strings corresponding to item information such as a title, document number, issue date, company name, and amount, from a group of character strings included in the document image sample 114. Thereafter, the learning device 102 transmits the generated machine learning model 115 to the information processing server 103 via the network 104.

[0015] The information processing server 103 is composed of an information processing unit 154 and a storage unit 155. The information processing unit 154 executes OCR processing (character recognition processing) on ​​the document image 113 received from the information processing device 101, and acquires a group of character strings included in the document image 113. Next, the information processing unit 154 uses the machine learning model 115 acquired from the learning device 102 to acquire a document type determination result 117 and an item information extraction result 118.

[0016] In this embodiment, the information processing unit 154 uses a document type determination model, which is a machine learning model 115 for determining the document type, to obtain a document type determination result 117 indicating whether the document image 113 is classified into a document type such as an invoice, an estimate, a purchase order, or a delivery note. The document type here is the type of a form. Furthermore, the information processing unit 154 uses an item information extraction model, which is a machine learning model 115 for extracting item information, to extract character strings corresponding to predetermined item information such as a title, a document number, an issue date, a company name, and an amount from a group of character strings included in the document image 113. The extracted character strings are obtained as an item information extraction result 118. The storage unit 155 stores a file of the document image 116 that is the target of processing by the information processing unit 154 among the document images 113 received from the information processing device 101. When storing the file of the document image 116, the storage unit 155 uses the character strings of the document type determination result 117 and the item information extraction result 118 as file names.

[0017] The network 104 is realized by a LAN, a WAN, or the like, and is a communication unit that connects the information processing device 101, the learning device 102, and the information processing server 103 to one another and transmits and receives data between the devices.

[0018] <Device configuration> FIG. 2 is a diagram showing an example of the configuration of an information processing device 101, a learning device 102, and an information processing server 103 for implementing the information processing system 100 in FIG.

[0019] Fig. 2(a) is a diagram showing the configuration of an information processing device 101. As shown in Fig. 2(a), the information processing device 101 is composed of a CPU 201, a ROM 202, a RAM 204, a printer device 205, a scanner device 206, a storage 208, an external interface 211, etc., which are connected to each other via a data bus 203.

[0020] The CPU 201 is a control unit for controlling the overall operation of the information processing device 101. The CPU 201 executes a startup program stored in a ROM 202 to start up the system of the information processing device 101, and executes a control program stored in a storage 208 to realize functions of the information processing device 101, such as printing, scanning, and faxing. The ROM 202 is realized by a non-volatile memory, and is a storage unit for storing a boot program for booting the information processing device 101. The data bus 203 is a communication unit for transmitting and receiving data between devices constituting the information processing device 101 . The RAM 204 is realized by a volatile memory, and is a storage unit used as a work memory when the CPU 201 executes a control program.

[0021] The printer device 205 is an image output device, and is a processing unit for printing and outputting a document image inside the information processing device 101 onto a storage medium such as paper. The scanner device 206 is an image input device and a processing unit for optically reading a storage medium such as paper on which characters, diagrams, and the like are printed, and acquiring the data as a document image. The document transport device 207 is realized by an ADF (Auto Document Feeder) or the like, and is a processing unit for detecting a document placed on a document table and transporting the detected document to the scanner device 206 one sheet at a time.

[0022] The storage 208 is realized by a hard disk drive (HDD) or the like, and is a storage unit for storing the above-mentioned control programs and document images. The input device 209 is realized by a touch panel, hard keys, or the like, and is a processing unit for accepting operation input to the information processing device 101 from the user. The display device 210 is realized by a liquid crystal display or the like, and is a display unit for displaying and outputting the setting screen of the information processing device 101 to the user. The external interface 211 connects the information processing device 101 and the network 104, and is an interface unit for receiving FAX data from a FAX transmitter (not shown) and transmitting document images to the information processing server 103.

[0023] 2(b) is a diagram showing the configuration of the learning device 102. As shown in FIG. 2(b), the learning device 102 is composed of a CPU 231, a ROM 232, a RAM 234, a storage 235, an input device 236, a display device 237, an external interface 238, and a GPU 239, which are connected to each other via a data bus 233.

[0024] The CPU 231 is a control unit for controlling the overall operation of the learning device 102. The CPU 231 executes a boot program stored in the ROM 232 to start up the system of the learning device 102, and executes a learning program stored in the storage 235 to generate a machine learning model 115 for determining document types and extracting item information. The ROM 232 is realized by a non-volatile memory, and is a storage unit for storing a boot program that starts up the learning device 102. The data bus 233 is a communication unit for transmitting and receiving data between the devices that make up the learning device 102 . The RAM 234 is realized by a volatile memory, and is a storage unit used as a work memory when the CPU 231 executes the learning program.

[0025] The storage 235 is realized by a hard disk drive (HDD) or the like, and is a storage unit for storing the learning programs and document images described above. The input device 236 is realized by a mouse, keyboard, etc., and is a processing unit for receiving operational input from an engineer managing the learning device 102. The display device 237 is realized by a liquid crystal display or the like, and is a display unit for displaying and outputting the setting screen of the learning device 102 to an engineer. The external interface 238 connects the learning device 102 and the network 104, and is an interface unit for receiving document image samples 114 from the outside and transmitting machine learning models 115 to the information processing server 103. The GPU 239 is a calculation unit configured with an image processing processor. The GPU 239 executes calculations for generating the machine learning model 115 from a group of character strings included in a given document image, for example, in accordance with a control command given from the CPU 231.

[0026] 2(c) is a diagram showing the configuration of the information processing server 103. As shown in FIG. 2(c), the information processing server 103 is composed of a CPU 261, a ROM 262, a RAM 264, a storage 265, an input device 266, a display device 267, and an external interface 268, which are connected to each other via a data bus 263.

[0027] The CPU 261 is a control unit for controlling the overall operation of the information processing server 103. The CPU 261 executes a boot program stored in a ROM 262 to start up the system of the information processing server 103, and executes an information processing program stored in a storage 265 to perform information processing such as character recognition (OCR) and information extraction. The ROM 262 is realized by a non-volatile memory, and is a storage unit for storing a boot program that starts up the information processing server 103. The data bus 263 is a communication unit for transmitting and receiving data between devices constituting the information processing server 103 . The RAM 264 is realized by a volatile memory, and is a storage unit used as a work memory when the CPU 261 executes an information processing program.

[0028] The storage 265 is realized by a HDD (Hard Disk Drive) or the like, and is a memory unit for storing the aforementioned information processing program, the machine learning model 115, the document image 116, the document type determination result 117, the item information extraction result 118, etc. The input device 266 is realized by a mouse, keyboard, etc., and is a processing unit for accepting operational input to the information processing server 103 from a user who uses the information processing server 103 or an engineer who manages the information processing server 103. The display device 267 is realized by an LCD display or the like, and is a display unit for displaying and outputting the setting screen of the information processing server 103 to a user who uses the information processing server 103 or an engineer who manages the information processing server 103. The external interface 268 connects the information processing server 103 and the network 104, and is an interface unit for receiving the machine learning model 115 from the learning device 102 and receiving the document image 113 from the information processing device 101.

[0029] <Usage sequence> FIG. 3 is a diagram showing a usage sequence of the information processing system 100 of FIG. Fig. 3(a) is a diagram explaining the flow in which the learning device 102 generates a document type determination model and an item information extraction model, which are machine learning models 115, in accordance with data and instructions provided by an engineer. Note that the detailed processing flow of Fig. 3(a) will be described later with reference to Fig. 4.

[0030] First, an engineer developing the information processing system 100 inputs a plurality of document image samples 114 (a group of document images) to the learning device 102 (S301). Next, the learning device 102 executes OCR processing on the plurality of document image samples 114 input in S301 to obtain a group of character strings, and obtains the document type of each of the document image samples 114 (e.g., invoice, estimate, purchase order, delivery note) as a correct label. The document type here is the type of form. For a single document image sample classified into a plurality of document types, such as an "invoice and delivery note," the learning device 102 may assign a correct label by assigning "1" to each of "invoice" and "delivery note" and "0" to other document types. The correct label may be assigned manually by an engineer, or may be assigned automatically using an output result obtained by applying a trained machine learning model to a document image sample given in advance.

[0031] Next, the learning device 102 uses the character strings included in the document image sample 114 and the correct label of the document type as learning data to train the machine learning model 115 and generate a document type determination model (S302). Next, the learning device 102 transmits the document type determination model generated in S302 to the information processing server 103 and stores it in the storage unit 155 of the information processing server 103 (S303).

[0032] Next, the learning device 102 acquires, as in S302, a character string corresponding to predetermined item information (e.g., title, document number, issue date, company name, amount) from a group of character strings included in the document image sample 114 as a correct label. Here, the correct label may be manually assigned by an engineer, or may be automatically assigned using an output result obtained by applying a trained machine learning model to a document image sample given in advance. Using the group of character strings that are the OCR results of the document image sample 114 and the correct labels of the item information as training data, the machine learning model 115 is trained to generate an item information extraction model (S304). Next, the learning device 102 transmits the item information extraction model generated in S304 to the information processing server 103, and stores it in the storage unit 155 of the information processing server 103 (S305).

[0033] Next, the information processing server 103 acquires the file storage rule setting input by the engineer via the input device 236 as the default setting for the information processing system 100 (S306). The file storage rule set here is applied when the user uses the information processing system 100 without setting a file storage rule in advance, and is updated by the file storage rule provided by the user, which will be described later. Thereafter, the operation of the sequence diagram in FIG. 3(a) ends.

[0034] Fig. 3(b) is a diagram for explaining a flow of assigning a file name to the document image 113 acquired by the information processing device 101 using the document type determination result 117 and the item information extraction result 118 of the document image 113, and storing the file of the document image 113 in a predetermined storage location. Note that the detailed processing flow of Fig. 3(b) will be described later with reference to Figs. 5 to 7.

[0035] First, the information processing server 103 acquires file storage rule settings as user custom settings of the information processing system 100 by the user's operation input of the input device 266 (S311). Next, the user who uses the information processing system 100 places the original 111 on the information processing device 101 and executes scanning (S312). In response to this, the information processing device 101 transmits the document image 113 acquired by scanning the original 111 to the information processing server 103 (S313). Next, the information processing server 103 acquires a group of character strings included in the document image 113 transmitted in S313, determines the document type of the document image 113 using the document type determination model acquired in S303, and acquires character strings corresponding to the determined document type (S314).

[0036] Furthermore, the information processing server 103 acquires the group of character strings included in the document image 113 transmitted in S313, and extracts a group of character strings corresponding to predetermined item information from the group of character strings included in the document image 113 using the item information extraction model acquired in S305 (S315). After that, the information processing server 103 generates a file of the document image 113, and stores the file in a predetermined storage location according to the file storage rules acquired in S306 and S311. At this time, the information processing server 103 assigns a file name using the character strings acquired in S314 and S315 (S316). After that, the operation of the sequence diagram in FIG. 2(b) ends.

[0037] <Machine learning model generation> 4 is a flowchart showing a flow in which the learning device 102 generates a document type determination model and an item information extraction model, which are the machine learning model 115. Note that an execution program for each step shown in FIG. 4 is stored in the ROM 232, the RAM 234, or the storage 235 of the learning device 102, and is executed by the CPU 231 or the GPU 239 of the learning device 102.

[0038] First, in S401, the CPU 231 acquires a plurality of document image samples 114 input by an engineer in S301 of Fig. 3. Specifically, the CPU 231 acquires document image samples 114 that are generally called semi-standard forms such as invoices, estimates, and purchase orders, and that are created with different layouts for each issuing company. Next, in S402, the CPU 231 executes block selection (BS) processing and OCR processing on the document image sample 114 acquired in S401, and acquires a group of character strings that constitute the document image sample 114 as the OCR result. The acquired group of character strings may be handled in units of word segments that are arranged by dividing the document image by blank spaces or lines, using a technique such as block selection (BS), which is a known technique for identifying object units that constitute a document image of a semi-standard form. The acquired group of character strings may also be handled in units of word segments that are divided by using a morphological analysis technique on a sentence included in a sentence image, for example.

[0039] Next, in S403, the CPU 231 assigns a first correct label indicating a document type, such as an invoice, an estimate, or a purchase order, to the document image sample 114 formed of the character string group acquired in S402. Next, in S404, the CPU 231 assigns a second correct label indicating that the character string to be extracted is item information to be extracted, to the character string to be extracted that is included in the character string group acquired in S402.

[0040] Next, in S405, the CPU 231 trains the machine learning model 115 for realizing document type determination of the document image and item information extraction using the group of character strings acquired in S402, the first correct label assigned in S403, and the second correct label assigned in S404. Here, the machine learning model 115 may use a feature vector indicating the feature amount of the character string converted using known techniques such as Word2Vec, fastText, BERT, XLNet, ALBERT, etc., and the position coordinates of the character string arranged in the document image. Specifically, by using a trained language model of BERT, a general sentence (e.g., the full text of an article in Wikipedia (registered trademark)) can be converted into a feature vector expressed by a 768-dimensional numerical value for each character string data. In addition, the machine learning model 115 trains so as to be able to output the correct label assigned in S403 and S404, and outputs a predetermined document type label and a predetermined item information label for the input group of character strings as an inference result by the learning model. Here, the learning model may be a commonly known machine learning algorithm such as logistic regression, decision tree, random forest, support vector machine, neural network, etc. Specifically, the model outputs an inference result of one of the predetermined document types and the label of the item information according to the output value of the fully connected layer of the neural network to which the feature vector output by the BERT language model is input.

[0041] Next, in S406 , the CPU 231 transmits the document type determination model, which is the machine learning model 115 trained in S405 , to the information processing server 103 and stores it in the storage unit 155 of the information processing server 103 . Furthermore, in S407, the CPU 231 transmits the item information extraction model, which is the machine learning model 115 trained in S405, to the information processing server 103 and stores it in the storage unit 155 of the information processing server 103. Thereafter, the processing of this flowchart ends. According to the above flowchart of Figure 4, a document type determination model and an item information extraction model, which are machine learning models 115, can be generated using multiple document image samples 114 and stored in the memory unit 155 of the information processing server 103.

[0042] <Storage of document image files> 5 is a flowchart showing a flow in which the information processing server 103 stores a link to a file of the document image 113, the file name of which is given using character strings of a document type determination result and an item information extraction result obtained by applying the machine learning model 115 to the document image 113. Note that an execution program for each step shown in FIG. 5 is stored in any one of the ROM 262, the RAM 264, or the storage 265 of the information processing server 103, and is executed by the CPU 261 of the information processing server 103.

[0043] First, in S501, the CPU 261 acquires the document type determination model, which is the machine learning model 115 transmitted from the learning device 102. Next, in S502, the CPU 261 acquires the item information extraction model, which is the machine learning model 115 transmitted from the learning device 102. Next, in S503, the CPU 261 acquires file storage rules that can be used in the information processing system 100 as the default settings given by the engineer in S306 or the user custom settings given by the user in S311.

[0044] Next, in S504, the CPU 261 acquires the document image 113 acquired by the information processing device 101. FIG. 8 shows a document image of an "invoice / delivery note" which is a document that combines multiple document types (hereinafter, referred to as a combined form) as a specific example of the document image 113. As shown in FIG. 8, the document image 113 includes character strings corresponding to item information such as a title 801, an issue date 802, a document number 803, a company name of the document issuer 804, and a total amount 805. Note that, in this embodiment, the "invoice / delivery note" will be described as an example of a combined form, but other combined forms such as an "estimate / purchase order", an "estimate / invoice", an "delivery note / receipt", and an "delivery note / receipt" can be similarly applied. In the following description, it is assumed that the document image 113 shown in FIG. 8 is acquired in S504.

[0045] Next, in S505, the CPU 261 executes the block selection (BS) processing and OCR processing described in S402 on the document image 113, and acquires a group of character strings included in the document image 113.

[0046] Next, in S506, the CPU 261 uses the document type determination model acquired in S501 to determine which document type the document image 113 is classified into. The document type here is the type of form. FIG. 9 is a diagram showing an example of a document type determination result. FIG. 9(a) shows an example in which a group of character strings acquired from a document image of "invoice" is input to the document type determination model. In this case, the CPU 261 uses the document type determination model to calculate a probability value (0 to 1) for each document type of invoice, estimate, purchase order, delivery note, and contract, and determines that the document type is "invoice" according to these probability values. Here, the document type determination result may be such that a document type with a probability value equal to or greater than a predetermined threshold value is output as the determination result. The threshold value is, for example, 0.9.

[0047] Fig. 9(b) shows an example in which a group of character strings obtained from a document image of "delivery note" is input to the document type determination model. In this case, similar to the case of Fig. 9(a), the CPU 261 calculates a probability value (0 to 1) for each document type, and outputs "delivery note", which is a document type with a probability value equal to or greater than a predetermined threshold. Fig. 9(c) shows an example in which a group of character strings obtained from a document image of "delivery note / invoice" is input to the document type determination model. In this case, similar to Fig. 9(a), the CPU 261 calculates a probability value (0 to 1) for each document type, and outputs "invoice" and "delivery note", which are document types with probability values ​​equal to or greater than a predetermined threshold value.

[0048] FIG. 10(a) is a diagram for explaining a flow of estimating and outputting a document type for an input of a group of character strings acquired from the document image 113 shown in FIG. 8 by a document type determination model, which is a machine learning model 115. As shown in FIG. 10(a), a group of character strings is input in units of character strings called tokens, arranged in order according to a predetermined reading order in the document image. Specifically, data is input together with commands such as [CLS] at the beginning of the character string and [SEP] at the delimiter. Here, the leading [CLS] can be treated as a feature vector indicating the features of the entire character strings constituting the document image, and a document type determination result can be obtained using the output result of a multi-class classifier to which the feature vector is input.

[0049] Next, in S507, the CPU 261 uses the item information extraction model acquired in S502 to extract a group of character strings corresponding to predetermined item information from the group of character strings included in the document image 113. FIG. 10(b) is a conceptual diagram illustrating a flow of converting the group of character strings included in the document image described above into a 768-dimensional feature vector using the above-mentioned BERT, and outputting any one of the labels for each character string unit. As shown in FIG. 10(b), the group of character strings is input in a string unit called a token, arranged in order according to a predetermined reading order in the document image. Specifically, data is input together with commands such as [CLS] at the beginning and [SEP] at the break point. Such input data is converted into a 768-dimensional feature vector (distributed representation) that represents the feature amount of the character string using the above-mentioned BERT, and the CPU 261 extracts candidate character strings having labels for each item using a multi-class classifier that inputs the converted feature vector.

[0050] FIG. 10(b) shows an example of the output result of a multi-class classifier using the BIO tag format used in the task of named entity recognition. The multi-class classification method using the BIO tag format is a method of expressing the range of candidate strings with the labels of each item by B (Bigin), I (Inside), and O (Outside). For example, the B-TITLE tag represents the first string of the title, the I-TITLE tag represents the string included in the string range of the title, and the O tag represents the string not classified into any label. According to a series of processes using the item information extraction model shown in FIG. 10(b), it is possible to obtain candidate strings corresponding to each item information (e.g., title, document number, publication date, company name, amount) from the string included in the document image. In the example of FIG. 10(b), the string range of I-TITLE following B-TITLE can be extracted as a candidate string for the title.

[0051] Next, in S508, the CPU 261 uses the document type determination result acquired in S506 and the character strings of the item information extraction result acquired in S507 to generate a file of the document image 113 to which a file name for classifying and storing the document image 113 has been assigned. In the example of Fig. 11(a), a file name 902 is assigned to the document image file 900 using the document type determination result "Invoice" and the item information extraction result character strings such as "20070805", "BN0037", "ABC Co., Ltd.", and "41,800 yen".

[0052] Next, in S509, CPU 261 determines whether the document type determination result acquired in S506 includes multiple document types (whether it is a dual-purpose form). If CPU 261 determines that the document type determination result acquired in S506 includes multiple document types (YES in S509), it acquires a document type other than the document type used in the file name in S508 from among the multiple document types, and proceeds to S510. On the other hand, if CPU 261 determines that a single document type is included (NO in S509), it proceeds to S511. Here, if a document type other than "invoice" ("delivery note") is included, such as "invoice / delivery note," then YES is determined in S509.

[0053] In S510, the CPU 261 generates a file link to the document image 113, to which a file name for classifying and storing the document image 113 is assigned, using the document type acquired in S509 and the character string of the item information extraction result acquired in S507. In the example of FIG. 11(a), a link 904 to a file 900 of the document image is shown. This link 904 is assigned a file name 903 using the document type "delivery note" other than "invoice" included in the document type determination result, and character strings such as "20070805", "BN0037", "ABC Co., Ltd.", and "41800 yen" that are the item information extraction result. In this way, the CPU 261 generates multiple file names 902, 903 corresponding to multiple document types included in the document type determination result for a single document image 113. That is, the CPU 261 generates multiple file names 902, 903 that are different for each of the multiple document types.

[0054] Next, in S511, the CPU 261 stores the document image file 900 generated in S508 with the file name 902 assigned, and the file link 904 generated in S510 with the file name 903 assigned, in a predetermined storage folder 901. Thereafter, the processing of this flowchart ends.

[0055] According to the flowchart in Fig. 5, when the document image 113 is a multi-purpose form, a file name for each document type can be given to the file of the document image 113 and the link to that file. A file name for each document type may be given to each of a plurality of file links. According to the example in Fig. 11(a), it is possible to search for the "delivery note / invoice" file in the storage folder 901 by using the keyword "invoice" or by using the keyword "delivery note".

[0056] As described above, according to the first embodiment, when storing a document image of a document (a dual-purpose form) that is a combination of multiple document types as a document file, it is possible to assign multiple file names corresponding to the multiple document types that constitute the dual-purpose form to a single document file. This can improve searchability when classifying and storing document files of dual-purpose forms. For example, after storing a document file of "delivery note and invoice" in a specified folder, whether searching as "delivery note" or "invoice", the desired document file can be specified using a character string such as the issue date or company name included in the file name. In other words, the effort and time required for file search can be reduced, and searchability can be improved.

[0057] [Variation 1] In the above embodiment, in the case of a dual-purpose form, a method is described in which a document file with a file name using one document type is stored, and a link to the document file with a file name using the other document type is stored. In this modified example, a method is described in which a copy of the document file is stored instead of a link to the document file.

[0058] Fig. 6 is a flowchart for explaining a flow of assigning a file name to a file copy of the document image 113 using character strings of a document type determination result and an item information extraction result obtained by applying the machine learning model 115 to the document image 113. Note that an execution program for each step shown in Fig. 6 is stored in any one of the ROM 262, the RAM 264, and the storage 265 of the information processing server 103, and is executed by the CPU 261 of the information processing server 103. The flowchart in Fig. 6 differs from the flowchart explained in Fig. 5 in that S610 to S611 are executed instead of S510 to S511. S610 to S611 will be explained below.

[0059] In S610, the CPU 261 generates a file copy of the document image 113 to which a file name for classifying and storing the document image 113 is assigned using the document type acquired in S509 and the character string of the item information extraction result acquired in S507. In the example of FIG. 11B, a copy 914 of the file 900 of the document image is shown. This copy 914 is assigned a file name 913 using the document type "delivery note" other than "invoice" included in the document type determination result, and character strings such as "20070805", "BN0037", "ABC Co., Ltd.", and "41800 yen" that are the item information extraction result. In this way, the CPU 261 generates multiple file names 902, 913 corresponding to multiple document types included in the document type determination result for a single document image 113. That is, the CPU 261 generates multiple file names 902, 913 that are different for each of the multiple document types.

[0060] Next, in S611, the CPU 261 stores in a predetermined storage folder 901 the document image file 900 generated in S508 with the file name 902 assigned thereto, and the file copy 914 generated in S610 with the file name 913 assigned thereto.

[0061] According to the first modification, when the document image 113 is a multi-purpose form, a file name for each document type can be given to the file of the document image 113 and each copy of that file. A file name for each document type may be given to each of the multiple file copies. According to the example of FIG. 11(b), it is possible to search for the "delivery note / invoice" file in the storage folder 901 using the keyword "invoice" or the keyword "delivery note".

[0062] [Variation 2] In the above embodiment, in the case of a multi-purpose form, a method is described in which a document file given a file name using one document type is stored, and a link to a document file given a file name using the other document type is stored. In this modified example, in the case of a multi-purpose form, instead of the method of giving a file name for each document type, a method is described in which metadata including a character string of a document type determination result and an item information extraction result is given to a document file.

[0063] Fig. 7 is a flowchart for explaining a flow of adding metadata to a file of the document image 113 by using character strings of a document type determination result and an item information extraction result obtained by applying the machine learning model 115 to the document image 113. Note that an execution program for each step shown in Fig. 7 is stored in any one of the ROM 262, the RAM 264, and the storage 265 of the information processing server 103, and is executed by the CPU 261 of the information processing server 103. The flowchart in Fig. 7 differs from the flowchart explained in Fig. 5 in that S710 to S711 are executed instead of S510 to S511. S710 to S711 will be explained below.

[0064] In S710, the CPU 261 generates metadata for classifying and storing the document image 113, using the document type acquired in S509 and the character string of the item information extraction result acquired in S507. In the example of Fig. 11(c), metadata 923 including "delivery note", which is a document type other than "invoice" included in the document type determination result, is assigned to the document image file 900. In this manner, the CPU 261 generates a file name 902 including a character string corresponding to one of the multiple document types included in the document type determination result, and generates metadata including a character string corresponding to the other document type.

[0065] In S711, the CPU 261 assigns the metadata 923 generated in S710 to the file 900 of the document image generated in S508 and given the file name 902, and stores the file in a predetermined storage folder 901.

[0066] In Fig. 9(c), metadata 923 is added while maintaining file name 902, but as shown in Fig. 9(d), CPU 261 may include in the file name a character string to be added as metadata 923. Specifically, file name 932 may be generated by including in the file name a character string "invoice_delivery note" obtained by integrating "invoice" and "delivery note", which are the document type determination results, in a portion of the file name that corresponds to the document type.

[0067] According to the second modification, when the document image 113 is a dual-purpose form, metadata indicating the document type can be added to the file of the document image 113. According to the example of Fig. 11(c), by using the metadata 923, it becomes possible to search for the file "delivery note / invoice" in the storage folder 901 by the keyword "invoice" or by the keyword "delivery note".

[0068] [Embodiment 2] In the first embodiment, a method for determining whether a document is a dual-purpose form using a document type determination model is described. In the present embodiment, a method for determining whether a document is a dual-purpose form using a character string in the title of a document image extracted using an item information extraction model is described. Below, a description of the same parts as in the first embodiment is omitted.

[0069] <Storage of document image files> The process flow of this embodiment is similar to that of the first embodiment shown in the flowchart of Fig. 5, so the description will be focused on the differences and the same parts as those of the first embodiment will be omitted. The description of S501 to S508 will be omitted because they are similar to those of the first embodiment.

[0070] In S509, the CPU 261 determines whether or not the document type determination result acquired in S506 includes multiple document types (whether or not it is a dual-purpose form). In the second embodiment, in order to deal with a case where the document type determination result does not properly recognize multiple document types, the title character string extracted using the item information extraction model is referenced. Specifically, as shown in FIG. 9(c), in order to deal with a case where a "delivery note" is output as the document type determination result for a "delivery note / invoice" document image 113 and it cannot be recognized as a dual-purpose form that also serves as an "invoice," the title character string of the document image 113 is referenced.

[0071] FIG. 12 is a diagram for explaining the relationship between a character string included in the title of a document image and a document type. FIG. 12(a) shows an example in which a title corresponding to both an "invoice" and a "delivery note" is written in a single location. As shown in FIG. 12(a), a title 1201 of a document image 1200 may be written as "invoice (delivery note)" and a title 1211 of a document image 1210 may be written as "delivery note and invoice". In this embodiment, the CPU 261 uses a specific character string (hereinafter referred to as a specific character string) serving as a keyword for each document type as shown in FIG. 12(c) to determine whether or not a specific character string corresponding to a plurality of document types is included in the character string of the title of the document image 113.

[0072] A specific example will be described. The character string of the title 1201 of the document image 1200 in FIG. 12(a) includes "billing" 1202, which is a specific character string 1230 of "invoice", and "delivery" 1203, which is a specific character string 1231 of "delivery note". Therefore, the CPU 201 determines that the document image 1200 is a dual-purpose form corresponding to both the "invoice" and the "delivery note". Similarly, the character string of the title 1211 of the document image 1210 in FIG. 12(a) includes "billing" 1213, which is a specific character string 1230 of "invoice", and "delivery" 1212, which is a specific character string 1231 of "delivery note". Therefore, the CPU 201 determines that the document image 1210 is a dual-purpose form corresponding to both the "invoice" and the "delivery note".

[0073] FIG. 12(b) shows an example in which a title corresponding to both an "invoice" and a "delivery note" is written in multiple locations. As shown in FIG. 12(b), a first title 1221 of a document image 1220 may be written as "delivery note" in one location, and a second title 1222 may be written as "invoice" in a location other than the "delivery note". The character strings of the titles 1221 and 1222 of the document image 1220 in FIG. 12(b) include "invoice" 1224, which is a specific character string 1230 of an "invoice", and "delivery" 1223, which is a specific character string 1231 of a "delivery note". Therefore, the CPU 201 determines that the document image 1220 is a dual-purpose form corresponding to both an "invoice" and a "delivery note", similar to the example in FIG. 12(a).

[0074] The above-described process of determining whether or not the document image 113 is a multipurpose form using the character string included in the title of the document image 113 may be executed if it is determined in S509 that the document type determination result acquired in S506 does not include multiple document types. Alternatively, without using the document type determination model, in S509 the CPU 261 may determine whether or not the document image 113 is a multipurpose form using the character string included in the title of the document image 113. In this case, S501 and S506 may be skipped.

[0075] As described above, according to the second embodiment, it is possible to determine whether a document is a multipurpose form by using the character string in the title of a document image extracted by using the item information extraction model.

[0076] [Embodiment 3] In the first embodiment, a method for assigning a file name using a character string resulting from extraction of item information that is common regardless of the document type in the case of a multi-purpose form was described. In this embodiment, a method for assigning a file name using a character string resulting from extraction of item information that differs for each document type is described. Below, a description of the same parts as in the first embodiment will be omitted.

[0077] <Storage of document image files> The process flow of this embodiment is similar to that of the first embodiment shown in the flowchart of Fig. 5, so the description will be focused on the differences and the same parts as those of the first embodiment will be omitted. The description of S501 to S503 will be omitted because they are similar to those of the first embodiment.

[0078] In S504, the CPU 261 acquires the document image 113 shown in Fig. 13 as the document image 113 acquired by the information processing device 101. Fig. 13 shows a document image of a "bill / delivery note" which is a dual-purpose form as another specific example of the document image 113 in Fig. 8 used in the description of the first embodiment. As shown in Fig. 13, the document image 113 includes character strings corresponding to item information such as a title 1301, a delivery date 1302, a delivery number 1303, an invoice date 1304, an invoice number 1305, a company name 1306 that issued the document, and a total amount 1307. In the following description, it is assumed that the document image 113 shown in Fig. 13 is acquired in S504.

[0079] Since steps S505 to S506 are the same as those in the first embodiment, the description thereof will be omitted. In S507, the CPU 261 uses the item information extraction model acquired in S502 to extract a group of character strings corresponding to predetermined item information from the group of character strings included in the document image 113. Specifically, character strings corresponding to the item information of the title 1301, delivery date 1302, delivery number 1303, invoice date 1304, invoice number 1305, document issuing company name 1306, and total amount 1307 shown in FIG.

[0080] Next, in S508, the CPU 261 uses the document type determination result acquired in S506 and the character strings of the item information extraction result acquired in S507 to generate a file of the document image 113 to which a file name for classifying and storing the document image 113 has been assigned. In this embodiment, the CPU 261 generates a file name using the character strings of the item information extraction result related to the document type, thereby assigning a file name using the character strings of the item information extraction result that differ for each document type.

[0081] Specifically, the CPU 261 selects a character string corresponding to item information that is highly related to the target document type from the character string group of the item information extraction result acquired in S507, and generates a file name that includes the selected character string. For example, the degree of relatedness between the document type and the item information is used. In the example of the document image 113 shown in FIG. 13, when the document type determination result "invoice" is the target, the CPU 261 selects character strings corresponding to the invoice date 1304 and invoice number 1305 as item information that is highly related to "invoice". Then, it generates a file name that combines the selected character string with character strings corresponding to the target document type "invoice" and the company name of the document issuer 1306 and the total amount 1307 as item information that is common regardless of the document type.

[0082] In the example of Figure 14(a), a file name 1404 is assigned to a document image file 1400 using the document type determination result "Invoice" and character strings such as "20070820," "SK0018," "ABC Co., Ltd.", and "41,800 yen," which are the results of item information extraction.

[0083] Since S509 is the same as in the first embodiment, the description will be omitted. In S510, the CPU 261 generates a file link to the document image 113, to which a file name for classifying and storing the document image 113 is assigned, using the document type acquired in S509 and the character string of the item information extraction result acquired in S507. In the example of the document image 113 shown in FIG. 13, when the CPU 261 targets a "delivery note", which is a document type other than a "bill", it selects character strings corresponding to the delivery date 1302 and the delivery number 1303, which are item information highly related to the "delivery note". Then, it generates a file name that combines the selected character string with character strings corresponding to the target document type, "delivery note", and the company name 1306 of the document issuer and the total amount 1307, which are item information common regardless of the document type.

[0084] 14(a), a link 1406 to a document image file 1400 is shown. This link 1406 is given a file name 1405 using character strings such as "delivery note", which is a document type determination result other than "invoice", and "20070805", "NH0037", "ABC Co., Ltd", and "41800 yen", which are item information extraction results.

[0085] Next, in S511, the CPU 261 stores in a prescribed folder the file 1400 of the document image generated in S508 and given the file name 1404, and the link 1406 to the file 1400 generated in S510 and given the file name 1405. In this embodiment, the CPU 261 stores the file 1400 and the link 1406 to the file 1400 in prescribed storage folders 1401 to 1403 by allocating them.

[0086] Specifically, when a folder is provided for each business partner, the CPU 261 selects a folder for the business partner using a character string corresponding to the business partner from the character strings of the item information extraction results acquired in S507. Then, the CPU 261 stores a document file and a file link in the selected folder. When a folder is provided for each document type, the CPU 261 selects a folder for the corresponding document type using the document type determination result acquired in S506, and stores a document file and a file link in the selected folder.

[0087] Note that Fig. 14(a) shows an example of generating the file link explained using Fig. 5, but in a similar manner, as shown in Fig. 14(b), a file copy explained using Fig. 6 may be generated. That is, in the same manner as the file name 1405 is given to the link 1406 to the document image file 1400, a file name 1408 may be given to the copy 1407 of the document image file 1400 shown in Fig. 14(b).

[0088] As described above, according to the third embodiment, when storing a document image of a document (a dual-purpose form) that is a dual-purpose form as a document file, it is possible to assign multiple file names corresponding to the multiple document types that constitute the dual-purpose form to a single document file. This can improve searchability when classifying and storing document files of dual-purpose forms. For example, when a document file of a "delivery note and invoice" is stored in a specified folder and then searched as a "delivery note", a desired document file can be specified using a character string related to the "delivery note" such as a delivery date or a delivery number included in the file name. Similarly, when a document file of a "delivery note and invoice" is stored in a specified folder and then searched as an "invoice", a desired document file can be specified using a character string related to the "invoice" such as an invoice date or an invoice number included in the file name. In other words, the effort and time required for file search can be reduced, and searchability can be improved.

[0089] Although the present invention has been described above with reference to the embodiments, the above embodiments are merely illustrative examples of the implementation of the present invention, and the technical scope of the present invention should not be interpreted as being limited by these embodiments. In other words, the present invention can be implemented in various forms without departing from its technical concept or main features.

[0090] (Other embodiments) The present invention can also be realized by a process in which a program for implementing one or more of the functions of the above-described embodiments is supplied to a system or device via a network or a storage medium, and one or more processors in a computer of the system or device read and execute the program. The present invention can also be realized by a circuit (e.g., ASIC) that implements one or more of the functions.

[0091] The disclosure of this embodiment includes the following configuration, method, and program. (Configuration 1) A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to item information from the document image; a storage means for storing the document image by giving the document image a file name; having An information processing device characterized in that, when the document image is classified into multiple document types by the classification means, the storage means assigns multiple different file names to the document image using the group of character strings extracted by the extraction means. (Configuration 2) The information processing device according to configuration 1, characterized in that when the classification means classifies the document image into a first type and a second type, the storage means assigns a first file name including a character string corresponding to the first type and a second file name including a character string corresponding to the second type. (Configuration 3) The information processing device according to configuration 2, wherein the storage means generates the first file name using a character string corresponding to the item information related to the first type, and generates the second file name using a character string corresponding to the item information related to the second type. (Configuration 4) The information processing device according to configuration 2 or 3, wherein the storage means stores a file of the document image given the first file name and a link to the file, and gives the second file name to the link. (Configuration 5) The information processing device according to any one of configurations 2 to 4, characterized in that the storage means stores a file of the document image given the first file name and a copy of the file, and gives the copy a second file name. (Configuration 6) The information processing device according to any one of configurations 1 to 5, wherein the classification means classifies the document type of the document image based on document types whose corresponding probabilities are equal to or greater than a threshold, the document types being output by inputting a group of character strings obtained from the document image into a first model. (Configuration 7) 7. The information processing apparatus according to any one of configurations 1 to 6, wherein the classification means classifies the document type of the document image by using a specific character string included in a character string of a title of the document image. (Configuration 8) The information processing device according to any one of configurations 1 to 7, wherein the extraction means extracts a group of character strings corresponding to the item information by inputting a group of character strings obtained from the document image into a second model. (Configuration 9) The information processing device according to any one of configurations 1 to 8, wherein the storage means uses the group of character strings extracted by the extraction means to select from a plurality of folders a folder for storing at least one of a file of the document image, a link to the file, and a copy of the file, each of which has been given the file name. (Configuration 10) 7. The information processing apparatus according to configuration 6, further comprising a learning means for learning the first model by using a group of document images to which a correct answer label indicating a document type has been added. (Configuration 11) 9. The information processing apparatus according to configuration 8, further comprising a learning means for learning the second model by using a group of document images in which a correct answer label is added to a character string corresponding to item information. (Configuration 12) The document image is an image obtained by scanning a form, 12. The information processing apparatus according to any one of configurations 1 to 11, wherein the document type is a form type. (Configuration 13) A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to the item information from the document image; a storage means for storing the document image by giving the document image a file name; having An information processing device characterized in that, when the classification means classifies the document image into multiple document types, the storage means generates metadata to be assigned to the document image using the group of character strings extracted by the extraction means. (Configuration 14) The information processing device according to configuration 13, characterized in that when the classification means classifies the document image into a first type and a second type, the storage means assigns a file name including a character string corresponding to the first type and a character string corresponding to the second type. (method) a classification step of classifying the document type of the document image; an extraction step of extracting a character string group corresponding to item information from the document image; a storing step of storing the document image by giving the document image a file name; Including, An information processing method characterized in that, when the document image is classified into multiple document types in the classification step, in the storage step, multiple different file names are assigned to the document image using the group of character strings extracted in the extraction step. (program) The computer of the information processing device, A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to the item information from the document image; a storage means for storing the document image by giving the document image a file name; Functioning as a When the classification means classifies the document image into a plurality of document types, the storage means assigns a plurality of different file names to the document image using the group of character strings extracted by the extraction means. [Explanation of symbols]

[0092] 100: Information processing system, 101: Information processing device, 102: Learning device, 103: Information processing server

Claims

1. A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to item information from the document image; a storage means for storing the document image by giving the document image a file name; having An information processing device characterized in that, when the document image is classified into multiple document types by the classification means, the storage means assigns multiple different file names to the document image using the group of character strings extracted by the extraction means.

2. 2. The information processing apparatus according to claim 1, wherein when the classification means classifies the document image into a first type and a second type, the storage means assigns a first file name including a character string corresponding to the first type and a second file name including a character string corresponding to the second type.

3. 3. The information processing device according to claim 2, wherein the storage means generates the first file name using a character string corresponding to the item information related to the first type, and generates the second file name using a character string corresponding to the item information related to the second type.

4. 3. The information processing apparatus according to claim 2, wherein said storage means stores a file of said document image given said first file name and a link to said file, and said link is given said second file name.

5. 3. The information processing apparatus according to claim 2, wherein said storage means stores a file of said document image given said first file name and a copy of said file, and gives said copy a second file name.

6. The information processing device according to claim 1, characterized in that the classification means classifies the document type of the document image based on document types whose probability of corresponding to each document type is equal to or greater than a threshold, which are output by inputting a group of character strings obtained from the document image into a first model.

7. 2. The information processing apparatus according to claim 1, wherein the classification means classifies the document type of the document image by using a specific character string included in a character string of a title of the document image.

8. The information processing apparatus according to claim 1 , wherein the extraction means extracts a group of character strings corresponding to the item information by inputting a group of character strings obtained from the document image to a second model.

9. The information processing device according to claim 1, characterized in that the storage means uses the group of character strings extracted by the extraction means to select from a plurality of folders a folder for storing at least one of a file of the document image, a link to the file, and a copy of the file, each of which has been given the file name.

10. 7. The information processing apparatus according to claim 6, further comprising: a learning unit that learns the first model by using a group of document images to which a correct answer label indicating a document type has been added.

11. 9. The information processing apparatus according to claim 8, further comprising: a learning unit that learns the second model by using a group of document images in which a correct answer label is added to a character string corresponding to item information.

12. The document image is an image obtained by scanning a form, 2. The information processing apparatus according to claim 1, wherein the document type is a form type.

13. A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to item information from the document image; a storage means for storing the document image by giving the document image a file name; having An information processing device characterized in that, when the classification means classifies the document image into multiple document types, the storage means generates metadata to be assigned to the document image using the group of character strings extracted by the extraction means.

14. 14. The information processing apparatus according to claim 13, wherein when the classification means classifies the document image into a first type and a second type, the storage means assigns a file name including a character string corresponding to the first type and a character string corresponding to the second type.

15. a classification step of classifying the document type of the document image; an extraction step of extracting a character string group corresponding to item information from the document image; a storing step of storing the document image by giving the document image a file name; Including, An information processing method characterized in that, when the document image is classified into multiple document types in the classification step, in the storage step, multiple different file names are assigned to the document image using the group of character strings extracted in the extraction step.

16. The computer of the information processing device, A classification means for classifying the document type of the document image; an extraction means for extracting a group of character strings corresponding to item information from the document image; a storage means for storing the document image by giving the document image a file name; Functioning as a When the classification means classifies the document image into a plurality of document types, the storage means assigns a plurality of different file names to the document image using the group of character strings extracted by the extraction means.

Images