JP7748337B2 - Document instruction extraction system, document instruction extraction method, and document instruction extraction program - Google Patents

Description

The present invention relates to technology for extracting and processing instructions contained in documents such as specifications for industrial product design.

To improve quality and reduce costs in the design and manufacturing of industrial products, design teams analyze past specifications to identify trends in product parts that have undergone repeated design changes, shape characteristics, and the types of changes. This analysis requires someone with product knowledge to read through all of the instructions in the specifications, interpret their meaning, and classify them, which takes a significant amount of time.

With regard to the analysis of documents in the medical field, there is known technology that targets medical observation documents such as electronic medical records and report systems, and identifies problems from free-form text using rules that define categories and keywords (see, for example, Patent Document 1).

Japanese Patent Application Laid-Open No. 2009-245232

For example, the method described in Patent Document 1 targets the medical field, constructing judgment rules consisting of categories and keywords for problems, and then sequentially editing and expanding them using an editor. In the medical field, medical terminology is relatively standardized across different medical settings, so this method makes it possible to expand a universal model.

On the other hand, in the fields of industrial product design and manufacturing, the shapes and component configurations of objects vary enormously, there are many proper nouns such as product models and component names, the design and manufacturing processes span many organizations, and the cycle of product change tends to be rapid. For this reason, the meaning of the instructions written in documents such as specifications may vary depending on the design site or product, and the same terms may be used with completely different meanings. Furthermore, because many organizations are involved, the structure of specifications and other documents may differ depending on the process, and the location and method of writing instructions in specifications may also vary widely. This can also be true for documents outside the fields of industrial product design and manufacturing.

The present invention was developed in consideration of the above circumstances, and its purpose is to provide technology that can properly extract instructions contained in a document and properly identify representative words that correspond to the words and phrases contained in the instructions.

To achieve the above-mentioned objective, one aspect of the intra-document instruction extraction system is an intra-document instruction extraction system that extracts instruction content from a document containing the instruction content and identifies a representative word that corresponds to a phrase included in the instruction content. The intra-document instruction extraction system has a processor and a memory unit. The memory unit stores one or more of the documents, instruction content identification information for identifying the instruction content from the document, and representative word identification information for identifying the representative word from the instruction content. The processor identifies the instruction content from the document based on the instruction content identification information, and identifies a representative word from the instruction content based on the representative word identification information.

This invention makes it possible to properly extract instructions contained in a document and properly identify representative words corresponding to the words contained in the instructions.

FIG. 1 is a diagram illustrating an example of a product according to the first embodiment. FIG. 2 is a diagram illustrating an example of a specification according to the first embodiment. FIG. 3 is a diagram showing the overall configuration of the document instruction extraction system according to the first embodiment. FIG. 4 is a functional configuration diagram of the document instruction extraction system according to the first embodiment. FIG. 5 is a diagram showing the configuration of the meta information cue word dictionary according to the first embodiment. FIG. 6 is a diagram showing the configuration of the instruction sentence cue word dictionary according to the first embodiment. FIG. 7 is a diagram showing the configuration of a representative word dictionary relating to a design object according to the first embodiment. FIG. 8 is a diagram showing the configuration of a representative word dictionary relating to CAD shapes according to the first embodiment. FIG. 9 is a diagram showing the configuration of a representative word dictionary relating to types of changes according to the first embodiment. FIG. 10 is a flowchart of the instruction sentence extraction and group word identification process according to the first embodiment. FIG. 11 is a flowchart of the meta information/instruction sentence extraction process according to the first embodiment. FIG. 12 is a flowchart of the directive statement division process according to the first embodiment. FIG. 13 is a flowchart of the group word specification process according to the first embodiment. FIG. 14 is a flowchart of the determination process according to the first embodiment. FIG. 15 is a flowchart of an example of a determination subprogram process according to the first embodiment. FIG. 16 is a flowchart of another example of the determination subprogram processing according to the first embodiment. FIG. 17 is a diagram showing an example of a determination result of a representative word related to a design object according to the first embodiment. FIG. 18 is a diagram showing an example of a determination result of a representative word related to a CAD shape according to the first embodiment. FIG. 19 is a diagram showing the configuration of an instruction content list according to the first embodiment. FIG. 20 is a diagram showing the layout of a summary table according to the first embodiment. FIG. 21 is a diagram showing the configuration of the tally screen according to the first embodiment. FIG. 22 is a functional configuration diagram of a document instruction extraction system according to the second embodiment. FIG. 23 is a diagram showing the configuration of a CAD model display screen according to the second embodiment.

Embodiments will be described with reference to the drawings. Note that the embodiments described below do not limit the invention as defined in the claims, and not all of the elements and combinations thereof described in the embodiments are necessarily essential to the solution of the invention.

In this embodiment, items in dictionaries and tables are enclosed in <>, and specific examples of data values are enclosed in " " and written in italics in figures.

(First embodiment)
Before describing the intra-document instruction extraction system 101, an example of a product that is the subject of a document (specification) used in the intra-document instruction extraction system 101, and a specification as an example of a document related to the product will be described.

Figure 1 is a diagram showing an example of the configuration of a product according to the first embodiment.

Product 300 is assembled from product body 301 and part 302 (also referred to as part A). Part 302 is attached with harness 303. In product 300, rib 307 is formed on the top of product body 301. A part mounting portion 305 to which part 302 is attached is formed on the front surface 304 of product body 301. In addition, a boss 306 is provided on the top of product body 301 for fixing harness 303 of part 302.

Figure 2 is a diagram illustrating an example of a specification document according to the first embodiment. The specification document in Figure 2 shows an example of a specification document that describes instructions (instructions) for design changes to the product 300 shown in Figure 1.

The specification 201 includes a main title display area 202, an issue date display area 203, a specification number display area 204, a product model display area 205, a change title display area 206, a directive description range 207, a summary display area 208, and a page display area 209.

The main title display area 202 contains the main title "Specifications." The issue date display area 203 contains the subtitle "Issue Date" and the issue date of the specifications 201, "2022/04/01." The specification number display area 204 contains the subtitle "Specifications Number" and the specification number "K1001." The product model display area 205 contains the subtitle "Product Model" and the model of the product 300, "KA1." The change title display area 206 contains the subtitle "Changes." The instruction statement area 207 contains one or more instruction statements corresponding to the changes to the design of the product 300. In the example of Figure 2, the instruction statement area 207 contains instruction statements indicating each instruction, listed in bullet points, with a specified first character at the beginning of each line. The deadline display area 208 displays the subtitle "Date" and the deadline for product 300, "2022/05/01." The page display area 209 displays the word "Page" and the page number, "1/2."

When it is desired to grasp trends in design changes based on multiple specifications such as those shown in Figure 2, the document instruction extraction system 101 identifies instruction sentences in the specifications and performs processing to identify and classify representative words that express the original meaning of the instructions from the instruction sentences.

In this process, simply extracting the words (characteristic words) contained in the instruction sentence and comparing them with a dictionary may not be enough to identify and classify the representative word that originally meant the instruction.

For example, if the instruction for a design change is "Expand the R of the upper part of the fixed part of part A," even if an attempt is made to identify the representative word for the part of the product that this instruction applies to, the words in the instruction do not contain information that the part is the front of the main body of the product that the instruction intends. Also, if the instruction is "Expand the R of the base of the harness boss of part A," even if an attempt is made to identify the representative word for the part of the product that this instruction applies to, the words in the instruction do not contain information that the part is the upper part of the main body that the instruction intends. Furthermore, although these two instruction statements contain the same word, "part A," the parts of product 300 that these instruction statements apply to are different.

Furthermore, the instruction "Enlarge upper R of fixed portion of part A" contains the word "upper portion," but this upper portion does not refer to the upper portion of the main body of product 300. Therefore, it is not possible to identify a representative word indicating the target portion of product 300 from the words and phrases in the instruction alone.

As another example, when trying to identify a representative word that represents the type of design change from a design change instruction, in the instruction "Enlarge the radius of the upper part of the fixing part of part A," "enlarge radius" refers to, for example, enlarging the convex radius of the fixing part of part A, whereas in the instruction "Enlarge the radius of the base of the harness boss of part A," it refers to enlarging the concave radius of the base of the boss. Even though the phrase "enlarge radius" is the same, these are different types of design changes, and it is not possible to directly identify a representative word that corresponds to the type of design change from the phrase in the instruction.

Furthermore, in the design of industrial products, minor changes to the model or the time of production can significantly change the relative positions of the product's parts or the shapes of certain parts. For example, boss 306 for securing harness 303 of part A of product 300 may be positioned on the bottom surface of product body 301 at a certain point in time. In this way, if the position of boss 306 is changed, the intended content of the instruction "Expand the base radius of the boss for part A's harness" must be recognized as a design change related to the bottom surface of the body, and a representative term must be specified.

In this way, in order to identify and classify the appropriate representative word from a directive sentence, it is necessary to recognize the original intended content based on the combination of words in the directive sentence and their order. It is also necessary to recognize that even if the same words are used, different intentions may be intended depending on other conditions.

The document instruction extraction system 101 takes into account the above-mentioned circumstances and performs processing to appropriately identify and classify representative words that express the original meaning of instructions from instruction sentences.

Next, we will explain the document instruction extraction system 101 in detail.

(Hardware configuration)
FIG. 3 is a diagram showing the overall configuration of the document instruction extraction system according to the first embodiment.

The document instruction extraction system 101 is composed of a computer, such as a PC (Personal Computer) or a general-purpose server. The document instruction extraction system 101 includes a CPU (Central Processing Unit) 11 (an example of a processor), a RAM (Random Access Memory) 12, a ROM (Read Only Memory) 13, a storage device 14, a communication interface (I/F) 15, a recording medium reader 17, an input device 18, and a display device 19. The CPU 11, RAM 12, ROM 13, storage device 14, communication I/F 15, recording medium reader 17, input device 18, and display device 19 are communicatively connected, for example, via a bus.

The CPU 11 executes programs stored in the RAM 12 and ROM 13 to perform various calculations and centrally control each part of the document instruction extraction system 101. The RAM 12 functions as a work area for storing the programs executed by the CPU 11 and necessary information. The ROM 13 stores the BIOS (Basic Input Output System).

The storage device 14 is an example of a storage unit and is a non-volatile storage device such as a hard disk drive (HDD) or solid state drive (SSD), which stores various information. In this embodiment, the storage device 14 stores an in-document instruction extraction program 20 and one or more document files 21. The document files 21 are, for example, files of documents (e.g., specifications) related to product design or manufacturing, and may include text data and images created using document editing software, or may be files created by performing OCR (Optical Character Recognition) on image data of specifications to convert characters in the image data into text data.

The communication I/F 15 is an interface for communicating with other devices via a network such as the Internet (not shown). The recording medium reader 17 is a device that reads data from a recording medium 16 such as a Blu-ray (registered trademark) Disc, a DVD (Digital Versatile Disc), or a CD (Compact Disc).

The input device 18 is, for example, a keyboard, mouse, etc., and accepts information input by the user. The display device 19 is, for example, a liquid crystal display, organic EL display, etc., and displays and outputs a user interface including various information.

In the document instruction extraction system 101, the document instruction extraction program 20 set up in the storage device 14 may be obtained by downloading it from a device on the Internet via the communication I/F 15, or it may be read by the recording medium reading device 17 from the recording medium 16 on which the document instruction extraction program 20 is recorded.

Note that while Figure 3 shows an example in which the intra-document instruction extraction system 101 is configured using a single computer, the present invention is not limited to this, and the intra-document instruction extraction system 101 may be configured using multiple server devices. When the intra-document instruction extraction system 101 is configured using multiple server devices in this manner, the intra-document instruction extraction program 20 may be executed in cooperation with the multiple server devices. In this case, the intra-document instruction extraction program 20 corresponds to a collection of programs that are distributed and set up across multiple server devices.

Next, we will explain the functional configuration of the document instruction extraction system 101.

(Functional configuration)
FIG. 4 is a functional configuration diagram of the document instruction extraction system according to the first embodiment.

The intra-document instruction extraction system 101 includes a dictionary data storage unit 107, a judgment subprogram storage unit 109, a file input unit 102, a text data extraction unit 103, a reference sentence extraction unit 104, a representative word identification unit 105, a counting unit 110, and an output unit 111. The dictionary data storage unit 107 and the judgment subprogram storage unit 109 are primarily configured by the RAM 12 and/or the storage device 14, while the file input unit 102, the text data extraction unit 103, the reference sentence extraction unit 104, the representative word identification unit 105, the counting unit 110, and the output unit 111 are primarily configured by the CPU 11 executing the intra-document instruction extraction program 20.

The dictionary data storage unit 107 stores a meta-information clue word dictionary D1, a demonstrative clue word dictionary D2, and one or more representative word dictionaries D3. Details of these dictionaries will be described later. The determination subprogram storage unit 109 stores one or more determination subprograms (an example of a determination program) for determining representative words. The determination subprograms are called and executed by the second determination unit 108.

The file input unit 102 reads the document file 21 to be processed from the storage device 14 and passes the document file data to the text data extraction unit 103. Note that the document file 21 may also be read from an external device via a network. The text data extraction unit 103 extracts text data from the data in the document file 21 and passes it to the instruction sentence extraction unit 104.

The instruction sentence extraction unit 104 extracts instruction sentences of the instruction content in the document file based on the text data of the document file, the meta information clue word dictionary D1, and the instruction sentence clue word dictionary D2, generates a data set including a list of instruction sentences (instruction sentence list) and meta information data, and passes this to the target word identification unit 105.

The representative word identification unit 105 has a first determination unit 106 and a second determination unit 108. The first determination unit 106 searches for characteristic words in instruction sentences based on the instruction sentence list and the representative word dictionary D3, and performs processing to identify candidate representative words. When it is necessary to execute a determination subprogram to determine whether a candidate representative word is a representative word, the second determination unit 108 executes the determination subprogram to determine whether the candidate is a representative word, and based on the result, generates a list of representative words (representative word list), and passes the representative word list, instruction sentence list, and meta information data to the aggregation unit 110.

The aggregation unit 110 passes a data set that compiles the meta information data, instruction sentence list, and representative word list, as well as data aggregated based on this data set (aggregated data: see Figure 20) to the output unit 111. The output unit 111 stores the passed data set and aggregated data in the storage device 14, and outputs to the display device 19 a screen showing an instruction content list (see Figure 19) that includes information from the data set, and a aggregation screen based on the aggregated data (see Figure 21).

(dictionary)
Next, the meta-information cue word dictionary D1, the instruction cue word dictionary D2, and the representative word dictionary D3 will be described in detail.

Figure 5 is a diagram showing the structure of the meta information clue word dictionary according to the first embodiment.

Meta information clue word dictionary D1 includes the entries <specification number start word>, <specification number format>, <specification date start word>, and <specification date format>.

<Specification Number Start Word> stores a word that indicates the start of the specification number in the specification. In the example of Figure 5, <Specification Number Start Word> stores "Specification Number", "Specification #", etc. <Specification Number Format> stores the format of the specification number. In the example of Figure 5, "A0001" is stored as the specification format. <Specification Date Start Word> stores a word that indicates the start of the specification date in the specification. In the example of Figure 5, <Specification Date Start Word> stores "Creation Date", "Registration Date", etc. <Specification Date Format> stores the format of the specification date. In the example of Figure 5, <Specification Date Format> stores "0000/00/00".

Figure 6 is a diagram showing the configuration of the instruction sentence cue word dictionary according to the first embodiment.

The instruction clue word dictionary D2 is an example of instruction content identification information and individual instruction identification information, and includes the following items: <specification start word>, <specification end word>, <instruction statement range start word>, <instruction statement range end word>, <line start character type>, <line number>, and <line start character>.

The <specification start word> stores a word that indicates the start position of the range in the specification to be processed. In the example in Figure 6, <specification start word> stores "specification", "specification document", etc. The <specification end word> stores a word that indicates the end position of the range in the specification to be processed. In the example in Figure 6, <specification end word> stores "page", etc.

The <start word of directive description range> field stores a word that indicates the start position of the range in the specification where the directive is written (directive description range). In the example in Figure 6, the <start word of directive description range> field stores "Change content." The <end word of directive description range> field stores a word that indicates the end position of the range in the specification where the directive is written. In the example in Figure 6, the <end word of directive description range> field stores "Essential."

<Line start character type> stores the character type that indicates the start of a line of an itemized directive within the directive description range. In the example in Figure 6, <Line start character type> stores "Num Type1", which indicates that the line start character is type [n], "Num Type2", which indicates that the line start character is type (n), etc. <Line number> stores the line number of the directive within the directive description range. In the example in Figure 6, <Line number> stores numbers from 1 to a predetermined number in order for each line start character type. <Line start character> stores the line start character that is indicated by the line number of the character line start type.

For example, in the example in Figure 6, the following are stored as "〈Line start character type〉,〈Line number〉,〈Line start character〉": "'Num Type1', '1', '[1]'", "'Num Type1', '2', '[2]'", ..., "'Num Type2', '1', '(1)'", "'Num Type2', '2', '(2)'", ....

In this embodiment, the representative word dictionary D3 (an example of representative word identification information) includes multiple representative word dictionaries for representative words of different categories: a representative word dictionary D3-1 related to the design object, a representative word dictionary D3-2 related to CAD shapes, and a representative word dictionary D3-3 related to the type of change.

Figure 7 is a diagram showing the configuration of the representative word dictionary for the design object in the first embodiment.

The representative word dictionary D3-1 is a dictionary for identifying representative words related to the design object (an example of a category), and includes items such as <process number>, <representative word>, <determination code>, <characteristic word (1)>, and <characteristic word (2)>. In the example of Figure 7, there are two items for storing characteristic words: <characteristic word (1)> and <characteristic word (2)>. However, if there are many characteristic words, there may be three or more items.

<Processing Number> stores the serial number (processing number) for the record in the target word dictionary D3-1. <Target Word> stores the target word candidate (target word). <Determination Code> stores determination process information related to the determination process used to determine whether the target word candidate for the <Target Word> corresponding to the record is the target word. The determination process information may include program-specific information that identifies the determination subprogram used in the determination process and information on the arguments passed to that determination subprogram. For example, the "no-" in "No-Part A, Part B" in the <Determination Code> of a record with a processing number of 2 is program-specific information that indicates the determination subprogram "Sub-n," and "Part A, Part B" are the arguments. Note that the determination subprogram "Sub-n" is a program that executes a determination process that determines, for example, whether a directive conforms to the condition that it does not contain a string corresponding to the argument. Additionally, the "date-after" in "date-after-20210501" in the <judgment code> of the record with processing number 4 is program-specific information that indicates the judgment subprogram "date-after," and "20210501" is the argument. Note that the judgment subprogram "date-after" is a program that executes judgment processing to determine whether a directive is satisfied because it is dated after the argument date. Note that if no data is set in the <judgment code>, the word in the <representative word> becomes the representative word without performing judgment processing.

<Feature Word (1)> and <Feature Word (2)> store one or more feature words that are the conditions for determining a record's representative word. In this embodiment, one of the conditions for a word in <Representative Word> to become a representative word is that the instruction statement contains a combination of one of the feature words in <Feature Word (1)> and one of the feature words in <Feature Word (2)>.

For example, in the fourth record in Figure 7, i.e., the record with a <processing number> of "4," the <representative word> stores "bottom of main body," the <determination code> stores "data-after-20210501," the <characteristic word (1)> stores "Part A," and the <characteristic word (2)> stores "harness, cord." This record indicates that if the combination of Part A and harness, or Part A and cord, is included in the directive, and the determination subprogram "date-after" indicated by "data-after" performs a determination process using "20210501" as an argument and determines that the combination is compatible, then the representative word is identified as "bottom of main body."

A specific product orientation can be expressed in a variety of ways, such as by expressions based on the orientation of the product in its normal state, by expressions based on the orientation of the product on design drawings, or by expressions based on the orientation during the manufacturing process. Therefore, even when the same word or phrase is used to indicate a direction, the actual meaning of the direction may be completely different. Therefore, the target word dictionary may include multiple correspondences for the same target word: a correspondence with a characteristic word (first characteristic word) related to the orientation of the product in its normal state, a correspondence with a characteristic word (second characteristic word) related to the orientation on the product's design drawings, and a correspondence with a characteristic word (third characteristic word) related to the orientation during manufacturing. For example, the target word dictionary may store at least multiple records: a record containing a characteristic word related to the orientation of the target word for the upper part of the main body in its normal state, a record containing a characteristic word related to the orientation of the upper part of the main body on the design drawings, and a record containing a characteristic word related to the orientation of the upper part of the main body during manufacturing. This makes it possible to identify the same target word regardless of the expression used to indicate the direction.

Figure 8 is a diagram showing the configuration of the representative word dictionary for CAD shapes in the first embodiment.

The representative word dictionary D3-2 is a dictionary for identifying representative words related to CAD shapes (an example of a category), and differs from the representative word dictionary D3-1 only in that the registered data relates to representative words related to CAD shapes; the item structure is the same.

Figure 9 is a diagram showing the configuration of the representative word dictionary for change types in the first embodiment.

The representative word dictionary D3-3 is a dictionary for identifying representative words related to change types (an example of a category). The only difference from the representative word dictionary D3-1 is that the registered data relates to representative words related to change types in product design content; the item structure is the same.

(Processing Operation)
Next, the processing operation of the document instruction extraction system 101 will be described.

(Instruction sentence extraction and representative word identification processing)
First, the instruction sentence extraction and group word specification process will be described.

Figure 10 is a flowchart of the instruction sentence extraction and group word identification process according to the first embodiment.

When the intra-document instruction extraction system 101 receives a trigger to start processing (e.g., a processing start instruction from the user via the input device 18) while in standby mode, the file input unit 102 accepts a selection from the user of one or more document files to be processed, reads the selected document files from the storage device 14, and passes the document files as data to the text data extraction unit 103 (S101).

Next, the text data extraction unit 103 extracts text data from the document file data and passes it to the instruction statement extraction unit 104 (S102).

Next, the instruction sentence extraction unit 104 executes a meta information/instruction sentence extraction process (see FIG. 11) to extract meta information of the document file and instruction sentences in the document file based on the text data of the document file, the meta information clue word dictionary D1, and the instruction sentence clue word dictionary D2, and passes a data set including a list of the extracted instruction sentences (instruction sentence list) and meta information data to the representative word identification unit 105 (S103).

Next, the target word identification unit 105 searches for characteristic words based on the instruction sentence list, meta information data, and target word dictionary D3, executes a target word identification process (see FIG. 13) to identify target words, generates a target word list, and passes the target word list, instruction sentence list, and meta information data to the aggregation unit 110 (S104). In this embodiment, the target word identification unit 105 performs the process of step S104 for all document files to be processed, and for each category of target words for each document file. For example, the target word identification unit 105 outputs the target word list, instruction sentence list, and meta information data obtained for all document files to be processed as a table-format file corresponding to the instruction content list R1 (see FIG. 19).

Next, the counting unit 110 reads the tabular file output from the target word identification unit 105, counts the number of target words identified for each target word category, creates a count table R2 (see Figure 20), stores it in the storage device 14, and outputs it as a tabular file (S105).

Next, the output unit 111 draws a graph showing the identified number of representative words for each representative word category based on the summary table R2, outputs this to the storage device 14 as an image file, and displays the summary screen 401 (see Figure 21) including the graph on the display device 19 (S106), thereby terminating the process.

(Meta information and instruction sentence extraction processing)
Next, the meta-information/instruction sentence extraction process in step S103 will be described.

Figure 11 is a flowchart of the meta information/instruction sentence extraction process according to the first embodiment.

First, the instruction sentence extraction unit 104 receives text data of a document file from the text data extraction unit 103 and stores that text data (S201).

Next, the instruction sentence extraction unit 104 extracts and stores the specification number from the text data based on the data in the <specification number start word> and <specification number format> in the meta information clue word dictionary D1 (S202).

Next, the instruction sentence extraction unit 104 extracts and stores the specification date from the text data based on the <specification date start word> and <specification date format> data in the meta information clue word dictionary D1 (S203).

Next, the instruction sentence extraction unit 104 extracts and stores the instruction sentence description range from the text data based on the data for the <instruction sentence description range start word> and <instruction sentence description range end word> in the instruction sentence clue word dictionary D2 (S204).

Next, the instruction sentence extraction unit 104 performs an instruction sentence division process (see FIG. 12) to divide the contents of the instruction sentence description range into individual instruction sentences based on the data for <first character type>, <line number>, and <first character> in the instruction sentence clue word dictionary D2 (S205).

Next, the instruction sentence extraction unit 104 stores the stored meta-information data such as the specification number and specification date, as well as the instruction sentence data list, as data available to the representative word identification unit 105 (S206), and then ends the process.

(Directive Sentence Splitting Processing)
Next, the instruction statement division process in step S205 will be described.

Figure 12 is a flowchart of the directive segmentation process according to the first embodiment.

When the instruction statement extraction unit 104 receives a trigger to start processing, it resets the value of the extracted flag to "False" and prepares an instruction statement list as an empty array variable (S211). Here, the instruction statement list has, for example, a pair of a number and a corresponding instruction statement as one element.

Next, the instruction sentence extraction unit 104 assigns the instruction sentence description range stored in step S204 to variable t1, and assigns the data of <first character type> in the instruction sentence cue word dictionary D2 to array variable L1 so as not to overlap (S212). As a result, array variable L1 stores all types of first character types registered in <first character type> in the instruction sentence cue word dictionary D2.

The directive extraction unit 104 then repeats the processing of loop A (S213 to S220) by assigning the values of array variable L1 one by one to variable n.

In the processing of Loop A, the instruction extraction unit 104 checks the value of the extracted flag (S213). If the extracted flag is "True" (S213: Yes), this indicates that the instruction has been split, and the process exits Loop A and ends.

On the other hand, if the extracted flag is "False" (S213: No), the instruction sentence extraction unit 104 assigns to array variable L2 all lists of pairs of line numbers in <line number> and line start characters in <line start character> that correspond to <line start character type> = variable n in the instruction sentence clue word dictionary D2 (S214).

Next, the directive extraction unit 104 repeats the processing of loop B (S215 to S220) by assigning the line number of <line number> to variable m and the first character of <line first character> to variable s, starting from <line number> = 1 for array variable L2.

In processing loop B, the directive extraction unit 104 searches for variable s within variable t1 and determines whether variable s is present within variable t1 (S215). As a result, if variable s is not present within variable t1 (S215: No), the directive extraction unit 104 determines whether the value of variable m is 1 (S221). If it is determined that variable m is 1 (S221: Yes), that is, if this is the first iteration of loop B, the directive extraction unit 104 exits loop B. On the other hand, if it is determined that variable m is not 1 (S221: No), that is, if this is the second or subsequent iteration of loop B, the directive extraction unit 104 adds variable t1 to the directive list as the (m-1)th directive (S220) and exits loop B.

On the other hand, if variable s is found in variable t1 (S215: Yes), the directive extraction unit 104 sets the position where variable s is found in variable t1 to p, assigns "True" to the extraction flag, and assigns everything from p to the end of variable t1 to variable t2 (S216).

The instruction sentence extraction unit 104 then determines whether the value of variable m is 1 (S217). If it determines that variable m is 1 (S217: Yes), i.e., if this is the first iteration of loop B, the processing proceeds to step S219. If it determines that variable m is not 1 (S217: No), i.e., if this is the second or subsequent iteration of loop B, the processing proceeds to step S218.

In step S218, the instruction statement extraction unit 104 adds the first instruction statement of variable t1 to p as the (m-1)th instruction statement to the instruction statement list, and proceeds to step S219. In step S219, the instruction statement extraction unit 104 assigns variable t2 to variable t1 (S219).

When the processing of step S219 is completed, the directive extraction unit 104 performs processing of loop B on the next line number, and when processing of loop B has been completed for all line numbers of the array variable, it exits loop B.

The directive extraction unit 104 then processes loop A, targeting the next value of array variable L1, and when loop A has been processed for all values of array variable L1, it exits loop A and terminates the directive splitting process.

(Representative word identification process)
Next, the representative word specification process in step S104 will be described.

Figure 13 is a flowchart of the target word identification process according to the first embodiment.

First, when the first determination unit 106 of the target word identification unit 105 receives a trigger to start processing, it stores the instruction sentence list generated by the instruction sentence extraction unit 104 (S301).

Then, the first determination unit 106 repeats the processing of loop C (S302 to S309) by assigning the values of the directive statement list one by one to variable t3.

In the processing of Loop C, the first judgment unit 106 repeats the processing of Loop D (S302 to S309) by obtaining the data for the representative word, second judgment code, characteristic word (1), and characteristic word (2) of one record in the representative word dictionary D3, starting from the top.

In the processing of Loop D, the first determination unit 106 divides the data for each of <Feature Word (1)> and <Feature Word (2)> at the comma ",", and creates keywords by combining and concatenating the character strings created by dividing the data for <Feature Word (1)> with the character strings created by dividing the data for <Feature Word (2)> in a brute-force manner, and generates a list of these keywords (keyword list) (S302). For example, if the target record is one with a <Processing Number> of 4 in Figure 7, a list containing the keywords "Part A Harness" and "Part A Code" will be generated.

Next, the first determination unit 106 searches for keywords in the keyword list using variable t3 and determines whether the keywords are found (S303).

As a result, if a keyword is found in variable t3 (S303: Yes), the first determination unit 106 proceeds to step S304, whereas if a keyword is not found in variable t3 (S303: No), the first determination unit 106 proceeds to step S308.

In step S304, the first determination unit 106 determines whether data is stored in the <determination code> of the record. As a result, if data is stored in the <determination code> (S304: Yes), the first determination unit 106 causes the second determination unit 108 to execute a determination process (see FIG. 14) based on the data contained in the <determination code> (S305).

Next, the first determination unit 106 receives the determination result from the determination process and determines whether the determination result is suitable as a representative word (S306). As a result, if the determination result is suitable as a representative word (S306: Yes), the first determination unit 106 proceeds to step S307, whereas if the determination result is not suitable as a representative word (S306: No), the first determination unit 106 proceeds to step S308.

If no data is stored in the <determination code> in step S304 (S304: No), or if the determination result in step S306 indicates that the record is suitable as a representative word (S306: Yes), the first determination unit 106 determines the character string in the <representative word> of the record as the representative word, associates it with variable t3 (instruction statement), and adds it to the representative word list (S307).

In step S308, the first judgment unit 106 determines whether there is a next record in the representative word dictionary. If there is no next record (S308: No), the representative word is added to the representative word list as "undetermined," whereas if there is a next record (S308: Yes), processing for this record is terminated.

When loop D processing for one record has finished, the first determination unit 106 performs loop D processing on the next record, and when loop D processing has been performed on all records, it exits loop D.

When loop D is exited, the first determination unit 106 processes loop C with the next value in the instruction statement list as the processing target, and when loop C has been processed with all values in the instruction statement list as the processing target, loop C is exited and the target word identification process is terminated.

(Determination process)
Next, the determination process in step S305 will be described.

Figure 14 is a flowchart of the determination process according to the first embodiment.

The second judgment unit 108 determines the judgment subprogram and arguments in the judgment subprogram to be used in the judgment process based on the judgment code (S401). For example, if the judgment code is the judgment code for a record with a <Processing Number> of 2 in Figure 7, it determines that the judgment subprogram to use is the judgment subprogram "Sub no" corresponding to "no-", and determines the arguments to be "Part A, Part B". Next, the second judgment unit 108 executes the determined judgment subprogram using the determined arguments, thereby performing the judgment subprogram process (see Figure 15) (S402). Next, the second judgment unit 108 returns the judgment result from the judgment subprogram process (S403) and ends the judgment process.

(Decision subprogram processing)
Next, an example of the determination subprogram process in step S402 will be described.

Figure 15 is a flowchart of an example of the judgment subprogram processing according to the first embodiment. The judgment subprogram processing in Figure 15 is realized by the second judgment unit 108 executing the judgment subprogram "Sub-n."

The second judgment unit 108, which executes the judgment subprogram "Sub-n," generates a list of keywords from the arguments (S501).

Next, the second determination unit 108 searches for keywords in the keyword list in variable t3 (instruction statement) and determines whether the keywords are found (S502).

If a keyword is found (S502: Yes), this means that the character string in the <target word> of the record being processed is not a target word, so the second judgment unit 108 determines the judgment result to be incompatible (S503) and proceeds to step S505.

On the other hand, if the keyword is not found (S502: No), this means that the character string in the <target word> of the record being processed is the target word, so the second judgment unit 108 determines the judgment result to be a match (S504) and proceeds to step S505.

In step S505, the second judgment unit 108 returns the judgment result and terminates the judgment subprogram processing.

Next, we will explain another example of the judgment subprogram processing in step S402.

Figure 16 is a flowchart of another example of the judgment subprogram processing according to the first embodiment. The judgment subprogram processing in Figure 16 is realized by the second judgment unit 108 executing the judgment subprogram "date-after."

The second judgment unit 108, which executes the judgment subprogram "date-after," acquires the argument as the reference date (S601).

Next, the second determination unit 108 compares the date of the specification containing the directive of variable t3 (specification date) with the reference date (S602) and determines whether the specification date is later than the reference date (S603).

As a result, if the specification date is later than the reference date (S603: Yes), this means that the character string in the <target word> of the record being processed is a target word, so the second judgment unit 108 determines the judgment result to be conforming (S604) and proceeds to step S606.

On the other hand, if the specification date is not later than the reference date (S603: No), this means that the character string in the <target word> of the record being processed is not a target word, so the second judgment unit 108 determines the judgment result to be non-conforming (S605) and proceeds to step S606.

In step S606, the second judgment unit 108 returns the judgment result and terminates the judgment subprogram processing.

Note that while two judgment subprograms, "Sub-n" and "date-after," have been described as examples of judgment subprograms, more judgment subprograms may be used. In this case, each judgment subprogram is stored in the judgment subprogram storage unit 109, and program-specific information that enables the second judgment unit 108 to identify the judgment subprogram is stored in the <judgment code>, and the second judgment unit 108 identifies and executes the judgment subprogram corresponding to the program-specific information. Furthermore, when adding a new judgment subprogram, the judgment subprogram is stored in the judgment subprogram storage unit 109, and the second judgment unit 108 is configured to be able to identify the new judgment subprogram from the judgment code. This makes it possible to easily determine whether a word is a representative word using a judgment subprogram that performs appropriate judgment processing.

(Results of the target word identification process)
Next, the determination result of the target word identification process for target words related to the design object will be described.

Figure 17 is a diagram showing an example of the results of determining representative words related to the design object according to the first embodiment. Figure 17 shows the results of determining representative words when the specification 201 shown in Figure 2 is used as the processing object and representative word identification processing is performed using the representative word dictionary D3 for representative words related to the design object shown in Figure 7.

The determination results shown in Figure 17 show each instruction sentence, the matching results for each record in the target word dictionary for each instruction sentence, and the target word identified as a result. Specifically, the instruction sentence "Enlarge R of upper fixing part of part A" matches the record with process number 1 in the target word dictionary, thereby identifying "front surface of main body" as the target word. Furthermore, the instruction sentence "Enlarge R of base of upper rib" matches the record with process number 2 in the target word dictionary, thereby identifying "top of main body" as the target word. Furthermore, the instruction sentence "Enlarge R of base of boss for part A harness" matches the record with process number 3 in the target word dictionary, thereby identifying "top of main body" as the target word.

Next, we will explain the results of the target word identification process for target words related to CAD shapes.

Figure 18 is a diagram showing an example of the results of determining representative words related to CAD shapes in the first embodiment. Figure 18 shows the results of determining representative words when the specification 201 shown in Figure 2 is processed and representative word identification processing is performed using the representative word dictionary D3 for representative words related to CAD shapes shown in Figure 8.

The determination results shown in Figure 18 show each instruction statement, the matching results for each record in the target word dictionary for each instruction statement, and the target word identified as a result. Specifically, the instruction statement "Enlarge R of upper fixed part of part A" matches the record with process number 3 in the target word dictionary, thereby identifying "other" as the target word. Furthermore, the instruction statement "Enlarge R of upper rib base" matches the record with process number 1 in the target word dictionary, thereby identifying "rib" as the target word. Furthermore, the instruction statement "Enlarge R of base of boss for part A harness" matches the record with process number 2 in the target word dictionary, thereby identifying "boss" as the target word.

(Instruction List)

Next, we will explain the instruction list R1.

Figure 19 is a diagram showing the structure of the instruction content list table according to the first embodiment.

Instruction content list R1 has the following fields: <Specification number>, <Creation date>, <Line number>, <Instruction statement>, <Design object>, <CAD shape>, and <Change type>.

<Specification number> stores the specification number. <Creation date> stores the date the specification was created. <Line number> stores the line number within the directive range of the directive corresponding to the line. <Directive> stores the directive corresponding to the line. <Design object> stores the representative word related to the design object in the directive corresponding to the line. <CAD shape> stores the representative word related to the CAD shape in the directive corresponding to the line. <Change type> stores the representative word related to the change type in the directive corresponding to the line.

(Summary table)
Next, the summary table R2 will be described.

Figure 20 is a diagram showing the layout of a summary table according to the first embodiment.

Summary table R2 has the following fields: <Instruction Category>, <Representative Word>, and <Number of Instructions>. <Instruction Category> stores the category of the representative word, such as design object, CAD shape, or change type. <Representative Word> stores the identified representative word. <Number of Instructions> stores the number of instructions for which the representative word is identified.

Next, we will explain the tally screen 401.

Figure 21 is a diagram showing the configuration of the tally screen in the first embodiment.

The summary screen 401 includes a design object graph display area 402 that displays a graph showing the number of identified representative words for the design object, a CAD shape graph display area 403 that displays a graph showing the number of identified representative words for the CAD shape, and a change type graph display area 404 that displays a graph showing the number of identified representative words for the change type.

The tally screen 401 allows you to easily see the identified representative words and their proportions for each category. This makes it easy to see which representative words are attracting attention in a document.

Second Embodiment
Next, a document instruction extraction system 101A according to a second embodiment will be described. Note that the same components as those in the document instruction extraction system 101 according to the first embodiment will be denoted by the same reference numerals.

(Functional configuration)
FIG. 22 is a functional configuration diagram of a document instruction extraction system according to the second embodiment.

The document instruction extraction system 101A further includes a product CAD model storage unit 112 and a CAD display unit 113 in addition to the document instruction extraction system 101.

The product CAD model storage unit 112 stores product CAD model information in which shape information (graphic data) of each object (component) that makes up the product is associated with a representative word that indicates each object. The CAD display unit 113 obtains, for example, an instruction content list R1 or a summary table R2 from the output unit 111, identifies objects associated with representative words that are frequently identified, and displays a CAD model display screen 505 (see FIG. 23) that includes a CAD model of the product in which the identified objects are emphasized (e.g., highlighted). For example, the representative word that is frequently identified may be the representative word that is identified the most, or it may be one or more representative words that are identified a number of times greater than a predetermined number.

Next, we will explain the CAD model display screen 505.

Figure 23 is a diagram showing the configuration of a CAD model display screen according to the second embodiment. Note that the CAD model display screen 505 in Figure 23 is the screen that is displayed when the boss 503 on the side of the main body is most frequently identified as the representative word.

The CAD model display screen 505 displays the CAD model 502 of the product, and highlights the boss 503 on the side of the main body that is most frequently identified as the representative word in the CAD model 502. This CAD model display screen 505 makes it easy to see which part is most frequently identified as the representative word, i.e., which part is most frequently specified in instruction sentences.

The present invention is not limited to the above-described embodiments, and can be modified and implemented as appropriate without departing from the spirit of the present invention.

For example, in the above embodiments, some or all of the processing performed by the CPU may be performed by hardware circuits. Furthermore, the programs in the above embodiments may be installed from a program source. The program source may be a program distribution server or a storage medium (e.g., a portable storage medium).

11...CPU, 14...storage device, 20...intra-document instruction extraction program, 21...document file, 101, 101A...intra-document instruction extraction system, D1...meta-information clue word dictionary, D2...reference sentence clue word dictionary, D3...representative word dictionary

Claims (12)

A document instruction extraction system that extracts instruction content from a document containing the instruction content and identifies a representative word that is a representative word corresponding to a phrase included in the instruction content,
The document instruction extraction system includes a processor and a storage unit.
The storage unit
one or more of said documents;
instruction content identification information for identifying the instruction content from the document;
and target word specification information for specifying the target word from the instruction content,
The processor:
Identifying the instruction content from the document based on the instruction content identification information;
Identifying a target word from the instruction content based on the target word identification information;
The representative word identification information corresponds to one or more characteristic words included in the instruction content, candidate representative words corresponding to the one or more characteristic words, and determination process information related to a determination process for determining that the candidate representative words are the representative words corresponding to the one or more characteristic words,
The processor:
determining whether the candidate for the target word is a target word based on the determination process information;
The storage unit
storing one or more determination programs for executing a determination process;
the determination process information includes information for identifying a determination program to be used in the determination process and information on arguments to be passed to the determination program;
The processor:
The determination program is executed using the arguments to determine whether the candidate representative word is a representative word.
A system for extracting instructions from documents. A document instruction extraction system that extracts instruction content from a document containing the instruction content and identifies a representative word that is a representative word corresponding to a phrase included in the instruction content,
The document instruction extraction system includes a processor and a storage unit.
The storage unit
one or more of said documents;
instruction content identification information for identifying the instruction content from the document;
and target word specification information for specifying the target word from the instruction content,
The processor:
Identifying instruction content from the document based on the instruction content identification information;
Identifying a target word from the instruction content based on the target word identification information;
For each target term identified from the plurality of documents, tallying the number of instructions in which the target term was identified;
Based on the aggregated results, create and display a graph showing the number of instructions for multiple representative words.
A system for extracting instructions from documents. A document instruction extraction system that extracts instruction content from a document containing the instruction content and identifies a representative word that is a representative word corresponding to a phrase included in the instruction content,
The document instruction extraction system includes a processor and a storage unit.
The storage unit
one or more of said documents;
instruction content identification information for identifying the instruction content from the document;
and target word specification information for specifying the target word from the instruction content,
The processor:
Identifying instruction content from the document based on the instruction content identification information;
Identifying a target word from the instruction content based on the target word identification information;
The storage unit
storing graphic data in which representative words are associated with components of the product;
The processor:
When displaying an image of a product based on the graphic data, a component of the product corresponding to a representative word identified from a document relating to the product is highlighted.
A system for extracting instructions from documents. A document instruction extraction system that extracts instruction content from a document containing the instruction content and identifies a representative word that is a representative word corresponding to a phrase included in the instruction content,
The document instruction extraction system includes a processor and a storage unit.
The storage unit
one or more of said documents;
instruction content identification information for identifying the instruction content from the document;
and target word specification information for specifying the target word from the instruction content,
The processor:
Identifying instruction content from the document based on the instruction content identification information;
Identifying a target word from the instruction content based on the target word identification information;
The document is a document related to the design and manufacturing of a specific product,
The target word specifying information is
For the same representative word, a plurality of correspondence relationships are included between at least a plurality of characteristic words among a first characteristic word relating to the direction of the product in normal times, a second characteristic word relating to the direction in design drawings of the product, and a third characteristic word relating to the direction of the product during manufacturing.
A system for extracting instructions from documents. A document instruction extraction method using a document instruction extraction system that extracts instruction content from a document containing the instruction content and extracts representative words that are representative words corresponding to words included in the instruction content,
The document instruction extraction system includes:
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
The representative word identification information corresponds to one or more characteristic words included in the instruction content, candidate representative words corresponding to the one or more characteristic words, and determination process information related to a determination process for determining that the candidate representative words are the representative words corresponding to the one or more characteristic words,
The document instruction extraction system includes:
determining whether the candidate for the target word is a target word based on the determination process information;
The document instruction extraction system includes:
storing one or more determination programs for executing a determination process;
the determination process information includes information for identifying a determination program to be used in the determination process and information on arguments to be passed to the determination program;
The document instruction extraction system includes:
The determination program is executed using the arguments to determine whether the candidate for the target word is the target word.
Intra-document instruction extraction method. A document instruction extraction method using a document instruction extraction system that extracts instruction content from a document containing the instruction content and extracts representative words that are representative words corresponding to words included in the instruction content,
The document instruction extraction system includes:
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
For each target term identified from the plurality of documents, tallying the number of instructions in which the target term was identified;
Based on the aggregated results, create and display a graph showing the number of instructions for multiple representative words.
Intra-document instruction extraction method. A document instruction extraction method using a document instruction extraction system that extracts instruction content from a document containing the instruction content and extracts representative words that are representative words corresponding to words included in the instruction content,
The document instruction extraction system includes:
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
When displaying an image of a product based on graphic data in which a representative word is associated with a component part of the product, the component part of the product corresponding to the representative word identified from a document relating to the product is highlighted.
Intra-document instruction extraction method. A document instruction extraction method using a document instruction extraction system that extracts instruction content from a document containing the instruction content and extracts representative words that are representative words corresponding to words included in the instruction content,
The document instruction extraction system includes:
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
The document is a document related to the design and manufacturing of a specific product,
The target word specifying information is
For the same representative word, a plurality of correspondence relationships are included between at least a plurality of characteristic words among a first characteristic word relating to the direction of the product in normal times, a second characteristic word relating to the direction in design drawings of the product, and a third characteristic word relating to the direction of the product during manufacturing.
Intra-document instruction extraction method. A document instruction extraction program executed by a computer to extract instruction contents from a document containing the instruction contents and extract representative words that are representative words corresponding to words included in the instruction contents,
The computer,
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
The representative word identification information corresponds to one or more characteristic words included in the instruction content, candidate representative words corresponding to the one or more characteristic words, and determination process information related to a determination process for determining that the candidate representative words are the representative words corresponding to the one or more characteristic words,
The computer,
determining whether the candidate representative word is a representative word based on the determination processing information;
the determination process information includes information for identifying a determination program to be used in the determination process and information on arguments to be passed to the determination program;
The computer,
The determination program is executed using the arguments to determine whether the candidate for the representative word is the representative word.
Document instruction extractor. A document instruction extraction program executed by a computer to extract instruction contents from a document containing the instruction contents and extract representative words that are representative words corresponding to words included in the instruction contents,
The computer,
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
For each target word identified from the plurality of documents, counting the number of instructions in which the target word was identified;
Based on the aggregated results, create and display a graph showing the number of instructions for multiple representative words.
Document instruction extractor. A document instruction extraction program executed by a computer to extract instruction contents from a document containing the instruction contents and extract representative words that are representative words corresponding to words included in the instruction contents,
The computer,
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
When displaying an image of a product based on graphic data in which a representative word is associated with a component part of the product, the component part of the product corresponding to the representative word identified from a document relating to the product is highlighted.
Document instruction extractor. A document instruction extraction program executed by a computer to extract instruction contents from a document containing the instruction contents and extract representative words that are representative words corresponding to words included in the instruction contents,
The computer,
Identifying the instruction content from the document based on instruction content identification information for identifying the instruction content from the document;
Identifying a target word from the instruction content based on target word identification information for identifying the target word from the instruction content;
The document is a document related to the design and manufacturing of a specific product,
The target word specifying information is
For the same representative word, a plurality of correspondence relationships are included between at least a plurality of characteristic words among a first characteristic word relating to the direction of the product in normal operation, a second characteristic word relating to the direction in the design drawing of the product, and a third characteristic word relating to the direction of the product during manufacturing.
Document instruction extractor.