JPH01306958A - Information expressing method and document processing system - Google Patents

Abstract

PURPOSE:To make duplex management of information in which documents are prepared by using paper for humans and floppy disks are used and mechanical codes of the documents are kept for machines unnecessary by preparing documents containing texts and graphics by using information which is visible to humans and information which is hardly visible to humans, but readable by machines. CONSTITUTION:A text 200 and graphic 210 which are visible by humans are described on a document 160 and, at the same time, information described with magnetized ink which is transparent or has the same color as the surface color of the paper 160 so that a machine can read the information is also recorded on the paper 160. Therefore, when, for example, the code 2-7 for machine is implicitly displayed on the display of the English word 'HOUSE' which is the visible information for humans by a concealing method, numerous codes for machine are repeatedly displayed in the area where the word 'HOUSE' is displayed. Even when, for example, the part 'HO' of the 'HOUSE' becomes stained, the word 'HOUSE' can be restored easily on a computer if even one of the codes for machine is left in a readable state in the remaining part. Therefore, duplex information management becomes unnecessary.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to the improvement of the interface between humans and computers, and in particular allows documents and drawings conventionally used in general society to be read not only by humans but also by computers. The present invention relates to a method of representing a document that facilitates and improves the interface by making it understandable, and a method of processing a document using the method of representation.

[Conventional technology]

Watermarks cannot be used for the purpose of the present invention because they can only be seen by humans, and the information expressed does not correspond to the information on the banknotes, or because they are created on paper in advance. Documents containing Braille for the blind are related to human vision and tactile sense2.
This is a simultaneous representation of multiple information, and is different from the dual information representation methods for humans and M11. A barcode encodes the selling price of a product, displays it next to the product name, and allows a calculator to read it directly, eliminating the hassle of inputting the selling price for payment calculations. The barcode is expressed in a visible format.

[Problem to be solved by the invention]

Conventionally, documents and graphical information created by computers have been stored or transmitted on floppy disks, magnetic cards, or magnetic tapes in a form that cannot be directly read by humans but can be read by computers. Separate display devices and line printer devices have been used to visually view the content. In this situation, it becomes necessary to manage the same information in a medium that is readable only by computers, and a medium that is readable by humans but difficult to read by computers. There is room for improvement.

As one direction for improvement, technology that allows computers to directly read human-readable information, such as text and graphics written on paper, has been researched and developed for several decades. However, research into character recognition or image recognition is generally said to be the most difficult of all research fields, and has only been put to practical use in very limited areas such as postal code classification. Even in this limited example, if the characters are smeared, it will immediately become impossible for a computer to do so. It is even more difficult to obtain graphic and image information, and the reality is that the development of automatic input of circuit drawings, that is, digitizers, is progressing slowly. The current conclusion of research in this direction is that unless a computer understands the meaning of the characters and figures it is trying to read, and by extension the meaning of the texts and drawings that contain them, full-fledged character and image recognition is impossible. It is said that it will be. Therefore, scanner technology that simply inputs information on paper as a bit pattern has begun to be used in personal computers, but it does not take into account the semantic information of the input object. Although a large-scale memory corresponding to the input bit plane is used, the image quality after input is significantly degraded due to noise, dirt on the input document, etc. Currently, there are high expectations for automatic translation between foreign languages. The ideal is to create an automatic translation machine that can create an automatically translated document that includes not only text but also graphics, etc., by simply providing a document that is commonly used in human society.

In order to make this possible, after solving the difficult problem of automatic document input mentioned above, we must solve two more difficult problems: document understanding and automatic translation, and it is expected that this will not become a reality in the distant future. be done.

The root of the problem lies in trying to force computers to imitate the functions of the human eye and brain, which have evolved over hundreds of millions of years, dating back to the origin of life.

On the computer side, since its origin, computer languages have developed to a certain degree, and with the development of data structure technology as a means of expressing the meaning of documents and figures in memory, computer processing of characters, 9 sentences, 9 figures, etc. is now easily possible. The situation has come.

Therefore, by mixing information representations in the same document that correspond to two entities, humans and computers, which have very different origins and internal processing mechanisms, it is possible to It is believed that communication between the parties will be greatly facilitated.

SUMMARY OF THE INVENTION An object of the present invention is to provide a document representation method that facilitates and improves the interface between humans and computers.

[Means to solve the problem]

In order to achieve the above object, the present invention includes conventional character and graphic information that is easily readable by humans but not necessarily easily readable by machines on a document, and a format that does not interfere with humans reading the above information. In other words, information written in a format that is difficult for humans to read is written on the same document, and the former is written when a human uses the document, and the latter is written when a machine processes the document. Try to look at it. In the current state of remarkable progress in miniaturized display technology, it is easy to write extremely large amounts of information for computers in an area that is visible to the human eye. Therefore, it is possible to express not only font information as a means of displaying characters and figures, but also the semantic structure of a single word or sentence, or even the semantic information of landscape drawings 1 and circuit drawings. By taking advantage of this, you can write the semantic structure information of the document on the document in advance, read it, and use it for translation.
It is also extremely easy to perform automatic translation from a document in one Japanese language to a document in another Japanese language as mentioned above.

In order to make the above basic idea a reality, it is necessary to develop a method for writing image information on a document, a method for creating a written explanation using a computer, or a method for reading it.

[Operation] In the present invention, an information code for human vision and an information code for machine reading are written side by side on the same document so that neither makes the other's information unreadable. By setting this to , humans can read documents as before, and machines can easily input documents through machine information codes. This method is
It is much easier and cheaper to realize this technology than conventional character and graphic recognition technology, which allows machines to directly read human visual information codes, and its performance is much higher. For example, the information code for human vision “Ho
When displaying the English word "use" by obscuring 2-7 ko 1-, the word "Mo
If a large number of machine codes are repeatedly displayed in the display rfI product of "use", even if the II HOI+ part becomes dirty with stains, as long as at least one machine code remains readable, The English word "Mouse" can be easily restored into the computer. This was not possible with the conventional character recognition technology described above.

It is common for the same document to contain a mixture of text and figures, and making a machine recognize these, including their meaning, is an even more difficult task than simple character recognition. Although research into making machines read the display directly has not yet been carried out, the method of the present invention allows the display origin 9 of the figure to be read by shadowing the figure display. Type of shape (circle, rectangle, etc.)
, size of figure, rotation angle 2 width of surrounding line, internal pattern 2
By displaying colors and other information using mechanical codes, figures can be easily read for practical purposes. Even if a part of the figure is dirty or cut, the original information can be easily read by using the above-described repeat display method.

Machines can easily interpret semantic information that is higher than the word level of a document by encoding it in the form of a parse tree and obscuring it on the document. Since the semantic structure of a sentence can be recognized, it becomes possible to easily convert the entire sentence into another sentence using this recognition. In other words, it is possible to easily translate a document into a foreign language. However, in this example, when the sentence is first created, the parse tree data is converted into a sentence in some way (for example, by adding a parse tree data creation function as part of the sentence creation editor's functions). A function is required to create a corresponding one.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

In FIG. 1, 100 and 100' are computer systems capable of outputting or inputting documents for simultaneous expression of different types of display means employing the present invention, 110 is a main body computer, 120 is a document input/output device, i 30 is a display, 140 is a mouse, and (1) 150 is a keyboard. The figure shows a situation in which a document 160 employing the representation method of the present invention is output by the computer system 100 and input by the computer system 100'.

FIG. 2(a) shows an example of a method for creating a document for simultaneous expression of different types of display means according to the present invention. The document in the figure: i 60 has human-visible text l-200 and a figure 210 written on it, but at the same time, for example, magnetic transparent or ink of the same color as the surface m'j of the paper 160. machine-readable information is also included.

This situation is expressed as A-A' section line 220 and B-B' section line 2.
30 (FIGS. 4(b) and (C)). In the cross-sectional view 240 of FIG. 2(b), a machine-readable visually non-readable material 250 is used below the display of the text "House" on the paper 242 using a visually readable material 252."T-101188 ” was created in 248 copies of 2.44,246° by repeating 7 times. The reason why it is displayed repeatedly here is
This is to ensure that even if part of the display becomes unreadable due to dirt, etc., the original information can be read. 2, a rectangular code representation 264 using machine-readable visually non-readable material 250 is shown in a cross-sectional view 260 along line BB′ of FIG.
゜It has been repeatedly created in 266 copies. Machine-readable visually-unreadable material information representation code begins”-F 270
, cord body portion 272. It is formed of a set of termination codes 274 to simplify machine reading. Low + in the diagram
:27o, 272, 274, etc. are simplified in illustration, but
In reality, it is assumed that information can be expressed sufficiently using a large number of bits.

FIG. 3 shows another embodiment of a simultaneous representation document with different display means. This method uses an encoding method, generally called MN encoding method, in which the frequency of bits 1 and 0 is relatively uniform among codes. Although it is a gIC, it has a machine-readable miniaturization level, and two cycles of code for machine characters II HI+ are written in copies 300 and 310. The code portion 340 is a non-information code portion. In each cycle, start code 320, 322, code body 325
327, and end codes 330 and 332 are displayed.

A cross-sectional view taken along the AA' cross-sectional line 360 is shown in FIG. 3(b). In this embodiment, information for humans and machines can be written on the paper surface 242 using only visually readable materials, so that the device 120 for inputting and outputting documents is different from the embodiment shown in FIG. can be realized at low cost.

FIGS. 4(a), (b), and (c) show a machine-readable text representation code 400. Figure table tax 1 code 430
9 shows the details of the sentence expression code 460.

Both have a code body between the start code 270 and the end code 274. The text expression code 400 indicates that the display object is text using the display object type specification code 410, then displays the display content, for example, "House" using the text specification code 412, and changes the color using the display color specification code 414. A font type designation code 416 indicates the type of display font, and a text coordinate code 418 indicates the coordinate point in the document of the display position of visible text on the document. lastly,
An error correction code 420 is added. Figure code 43
0, the display object type specification code 4S31 indicates that the display object is a figure, the figure type specification code 432 specifies whether the display figure is a rectangle or a circle, and the display color specification row +:434 specifies whether the display figure is a rectangle or circle. As for the color of the figure, the width of the outer boundary line is displayed by the figure outer width designation code 436, and the coordinates of the display position of the figure on the document are displayed by the figure coordinate designation code 438. Further, a figure meaning designation row 1-; 440 displays information such that the figure represents, for example, a wall of a house.

In the example of the text expression code 460, the entire document is, for example, “T
hj, s :is a House,” after displaying the text with 462, the verb of this sentence,
The subject and object are each displayed using a code of 464°466.4-68. The example shown here is only one example of various other possible code representation methods, and includes characters, words, 2 sentences, 9
Putting the shapes together (]5) Any method can be used as long as it expresses various objects in code.

Note that in the above explanation, the case where human-visible information such as "Th1s is a House" and the corresponding machine-readable information are expressed in overlapping positions on the document was shown, but both are not necessarily displayed in overlapping positions, and both may be displayed in separate positions.

FIG. 5 shows a document output and input processing mechanism 500 of the present invention.
This is an example. A computer 110 having a document processing function is connected to an input/output mechanism 500 forming a document input/output device 120 via a signal line 520. Input/output mechanism 50
0 is connected to the input/output head 540 via a signal line 530.

550 is a head moving arm. The head is provided with a knob 95602 for ejecting magnetic ink used when creating a document, a nozzle 5 f02 for ejecting normal ink, and a magnetic sensor used when inputting a document. The head 540 scans the scan belt 5 on the document 150.
90 in one stroke, and in the output mode outputs a machine-readable code with magnetically-containing ink and visually readable information with regular ink onto the document using nozzles 560 and 570, respectively. Figure 2 Creates a document for the embodiment.

In the input mode, the magnetic sensor 580 is used to read the machine readable code on the document. In the computer 110 with a document processing function, a document containing text and figures is created using document editing software i-ware, and its higher-order information code, for example, an ASCII code representation of the text, is
The input/output processing mechanism 50 is connected to the input/output processing mechanism 50 via the signal line 520 as one recorder sandwiched between the start mark and the end mark.
0 can be communicated. The input/output processing mechanism 500 uses a conversion dictionary to convert each element of this high-level information record, such as the - character in the case of text, into low-level information row: ( for printer output. At this time, visually readable information and A code paired with machine-readable information is created to control the printer head 540 and print image information on the document.When reading the document, machine-readable low-order information obtained by a magnetic sensor is used to control the printer head 540 and print image information on the document. is converted into high-order information using a dictionary by the input/output processing mechanism 500 and sent to the computer 110.A printer head 595 is shown when the embodiment shown in FIG. 3 is used as a method for writing information on a document. In this case, the ink nozzle 570 and the optical sensor 598 are usually used.
and on the head, when printing visible information,
Normally, by controlling on/off ink ejection from an ink nozzle in accordance with the bit pattern arrangement of the machine-readable code, it is possible to create visually visible information in which the MN code is hidden on the document. Conversely, when reading a document, the optical sensor 598 reads the MN code as low-order information, the input/output processing mechanism 500 converts it into high-order information, and transfers this to the computer 110.

FIG. 6 is a functional block diagram of the input/output processing mechanism 500. It is connected to the computer 110 via a high-order information path 520, and to the printer side via a low-order information path 530. At the time of output, the high-order information is temporarily stored in the data buffer 612. The high-order to low-order conversion dictionary sent from the computer 110 is stored in the dictionary memory 610. The high-order to low-order conversion unit 614 uses the dictionary in the dictionary memory 610 to convert the high-order information on the data buffer 612 to low-order information, and stores it in the low-order information data buffer 616. 6
After a parity code and a code are attached to the print control format 1 by the parity addition and format generation unit 617, the low-order information in the output data generation unit 6
19 into a print signal, and the D/A converter 618
is sent to the printer through

When a document is manually processed, an analog signal of machine-readable low-order information on the document is first digitized by an A/D converter 620, subjected to error correction by an error correction section 621, and then stored in a data buffer 622.

Since all the machine-readable codes expressed multiplexed on the document are read into the correction unit 612, the duplications are deleted by the duplication deletion processing unit 624, and the data from which the duplications have been removed is stored in the temporary memory 626. Thereafter, a low-order to high-order conversion unit 628 using a dictionary performs low-order to high-order conversion processing, and the obtained high-order information is stored in a data buffer 630 and sent to the computer 110.

In the figure, in order to explain output processing and input processing in an easy-to-understand manner, processing blocks that can be shared by both input and output processing are shown separately. , low-order information data buffer 6
16 and 622 may be the same. In addition, each part 614゜616, 617, 619, 621, 62
4,624, etc. can be implemented with one or more microprocessors.

FIG. 7 shows an overview of document editing processing on the computer 110 side. After processing mode determination 700, if the user's mode selection is document editing, document editing software 710 is activated to interactively create text and figures. At this time, the created text 1 and graphics store various information such as the display position on the document and the properties of the text and graphics in a high-level representation format such as ASCII code on the memory in the computer 110. Also, a set of information blocks, such as “House
” or for each shape such as a rectangle, a start mark and an end mark are attached to the higher-level information, and it is used as a unit record for subsequent output processing.If you want to perform translation etc., one sentence The entire sentence is defined as a unit record, and the syntax tree information of the sentence, such as the subject, a-word, object, etc., is defined in the record along with visual information.

If the user selects the output processing mode, process 720
will be carried out. That is, first, a dictionary for converting high-order information into low-order information is output to the input/output processing mechanism 500, and then the high-order information record on the computer memory is transferred to the data buffer 612 in FIG. Diagram processing units 614 to 61
8 performs printing processing.

When the user selects the input processing mode, the conversion dictionary is first transferred to the input/output processing mechanism 500.

Next, after reading the low-order information and converting it into high-order information by the processing units 620 to 628 in FIG. 6, all records in the data buffer 630 are taken into the computer memory.

FIG. 8 is an explanation of the contents of the conversion dictionary that is transferred to the dictionary memory 610 in FIG. 6 to perform conversion between high-order and low-order information. The header part 800 shows the field contents when the dictionary is placed in memory, the part 802 is the address, and the part 804 is the dictionary type. This is the part to do. The field 806 is a higher-order information field in which, for example, an ASCII code is written. Part 808 is the knee 1 of low-order visual information for each high-order information, part 810 is an invisible machine-readable information code, and the entire part 812 is a field for writing the low-order information code.

In the figure, part 814 is the dictionary content regarding the character 11 A +1. To convert certain high-order information to low-order information, match it with the bit pattern of field 806 in the figure,
All you have to do is extract the bit pattern in the low-level information field of the row that matches. Conversely, to convert machine-readable low-order information into high-order information, it is sufficient to match the bit pattern of code 810 in the figure and extract the contents of the high-order information field of the matching row.

FIG. 9 is a detailed explanation of the conversion process from high-order information to low-order information when outputting a document. data buffer 612
The records in 900 copies of the memory in the high-order → low-order conversion unit 614 are taken out for each unit information that constitutes the record, and are stored in 804 copies of the dictionary memory 610. '806 part is matched (processing 912), and the bit pattern of the low-order information part 808.810 of the matched row 914 is
(process 916) and transferred to the low-order information data buffer 616. As shown in the flowchart, first, it is checked whether the data buffer 6]2 is empty (processing 930), and if it is not empty, the low record start code is sent to the data buffer 616.
932), and then the data buffer 61
2, and in the case of a text record, its constituent unit, for example, a text record, is converted into low-level information using a dictionary, and then sent to the data buffer 61 one after another.
6 (processing 940). After converting one record, the end code of the lower record is sent to the data buffer 6.
16, a single low-order information record surrounded by a start code and an end code is created in the data buffer 616 (step 942). Finally, an output format instruction code is written into one low-order information record created in this way (processing 944). This can be, for example, visual information, “Hou
This is to specify the number of times the machine-readable code to be outputted is duplicated in se I+. After all records in the data buffer 612 are processed in the same manner, the process ends.

FIGS. 10(,) and (b) show details of the process of converting low-level information into high-level information when inputting a document. data buffer 6
The low-level information records surrounded by the start code and end code in 22 are taken into the duplicate data deletion unit 624 one by one (process 1000). Records with missing start codes or end codes are removed from the paper due to stains, etc. The information is considered to be corrupted and discarded (processing 1006). For other records, identity determination processing 1 is performed between the records 1 and 1002 stored in the -time memory 626.
004, and if the data is the same, it is considered to be duplicate data and discarded (processing 1006).

If they are not the same, store them in temporary memory (processing 1)
008). To determine the identity, perform an exclusive OR (FOR) operation between the bit pattern of record 1001 and the bit pattern of record 1002 for each bit, and if all the bits are 0, the data is the same. The method of determining this is simple. Note that, as a prerequisite, it is necessary to clear the temporary memory 626 before inputting a document.

The same content from different locations in the same document, for example "House"
” is input, the input low-level record also includes a display position coordinate code 418 as shown in FIG. 4, so identity is determined by EOR operation. Therefore, the duplicate record deletion process described above deletes only one piece of visible information displayed at the same location on the document when the document is dirty, etc. Only the redundant pieces of machine-readable information that were intentionally written in parallel in order to improve performance will be stored.When all the records in the data buffer 622 are processed and the data buffer 622 becomes empty, the records in the temporary memory will be are selected one by one, and are read into the memory in the conversion unit 628 one by one for each unit data array (1010, 1011, etc.) on each record (processing 1012), and this is stored in the lower-level information field in the dictionary. (process 1.020), and the bit patterns of the higher-order information fields 804 and 806 of the matched row 1022 are taken out to the memory 1030.

Next, the contents of the memory 1030 are transferred to the higher-order information data buffer 6.
Send it to 30. Note that before transferring one record to the data buffer 630, a record start code is sent to the data buffer 630, and when all processing of one record is completed, an end code is sent, and one record is transferred to the data buffer 630. Create higher-level information records. The above is explained as a flowchart in FIG. 10'(b).

As a result, the machine-readable information on the input document has been read into the data buffer 630 in the form of higher-order information records without duplication.

FIG. 11 explains the relationship between high-order information and low-order information. In this example, the English text "Th1s js a House" is used as a unit block of higher-order information. Of course, an entire sentence consisting of multiple sentences can be used as a unit block, but the h-method of definition can be arbitrarily determined depending on the convenience of document editing, and the example shown here is an absolute one. It's not a thing. Here, the unit block of higher-order information is several unit records 1
102, and the unit record has a high and number information body part 1112 between the start code ]-110 and the end code 1114, and the body part 1112 holds some intra-record unit information 11o3. I am giving an example of what is happening. A low-order information record corresponding to the high-order information record 1102 is shown at 1104. Starting low +: 1
120 and the end code 1130, an output format code 1122. Code body part 1124. It has a parity code of 1128. The unit information 1106 in the code body section 1124 is defined, for example, as a pair of a visible information code 1132 and a machine-readable information code 1134 for one character of text. Lower information unit record 11
04 is a unit block 11 of lower-order information that is assembled in several pieces.
08 is formed.

Next, the contents of the processing of the output data generation section 619 shown in FIG. 6 will be explained in detail with reference to FIG. 12. First, the low-level information record in the data buffer 616, with parity and output format information added, is received as input from the processing mechanism 617 (process 1200), and then this input information is processed. The lower information main body part (for example, the first
A bit pattern of text and figures to be output is created based on the contents of 1124 in FIG. 1 and the contents of the format code (for example, 1122 in FIG. 11) (process 1202).

Subsequently, the created bit pattern is output to the D/A converter line by line of the printer (process 1204). Processing 12
In step 02, both visual output information and mechanical output information are converted into bit patterns, and both are output in the output of process 1204. This output signal controls the operation of printing head 540 or 595 in FIG. 5 to display both visible and machine readable information on the same document.

FIG. 13 shows a case where a partially soiled document is input and edited as an applied example of the present invention. It is assumed that one copy of document 1300 (document 1) based on the present invention is stained.

Separately, a document 1304 (document 2) is also used. First, machine readable information on both documents is scanned and input (process 130).
6). Even if dirt 13o2 exists, the original information can be restored as long as there is one clean piece of information using the machine-readable information duplication output display method of the present invention. The input low-order information is converted to high-order information (processing 1308),
The information is imported into the main memory of the computer and edited using document editing software (processing 1310).

Since the input low-order information includes display position coordinates of texts and figures on the document, low-order → high-order conversion processing 13
If you convert that information into higher-order information in step 08, for example, you can convert the drawing of the sun in document 2 to document 1 through interactive processing.
It is easy to move it within the coordinate system of In addition, changes such as changing the character type in the document 1 and coloring the roof of a house shape in black can be easily accomplished through dialogue. After the document editing is completed, the high-level information of the new document created on the computer memory is converted to low-level information [processing 1312].
When the low-level information is output and displayed by the printer device (process 1.314), a new document [3]6 in which visually visible information and machine-readable information are written side by side is obtained. The major feature of the present invention compared to the prior art, as demonstrated by this application example, is that even if a document in which one part is damaged due to dirt or the like is input directly, the original information can be easily restored, and the original information can be easily restored. Direct reading of general documents, including the meaning of the information written on the documents, can be easily realized, and there is no need for dual ownership and management of floppy disks, etc. and printing paper as in the past. be.

Scanners, which are now widely used, can only input documents as simple black and white dot patterns, and it is impossible to restore and input the semantic information of text and figures. It is not possible to realize the advanced editing functions that can be achieved.

FIG. 14 is an explanatory diagram of a direct automatic translation method between documents in different languages to which the present invention is applied. In this example, original document 1
400, the semantic structure of the visible information is written in a machine-readable code along with the visible information in the form of a predicate used in logic. Easy home (red) by input processing.

The predicate expression of the sentence "Red House" can be restored on the computer's internal memory as higher-order information (process 1402).
). Each element of the predicate thus restored, i.e.
The predicate ``house'' and ``red'' are translated using the translation dictionary 1406 and written in another language, ``house(red'').

get.

Next, by outputting this newly obtained high-order information, it is possible to print out a new document 1408 having the foreign language text "Red house".

The significance of the present invention shown in this application example is that it is possible to easily directly convert a general document containing figures and text into a document in a different language, which has been thought to be impossible in the past. In traditional language translation, visible text,
In other words, in the case of this example, the information described in a certain review word, ``red house,'' is first converted into the predicate ``house (red)'' through syntactic analysis, and then this is converted into a foreign language. The parsing stage is an extremely difficult process, and it is not yet possible to perform a correct translation, and it is predicted that it will not be possible in the future. With current technology, it is easy to extract the true meaning structure of the sentence intended by the document creator through dialogue with the document creator during the process, write it in the form of predicates, etc., and store it separately from the sentence itself. In this application example, the explanation will be made assuming that the predicate (red)'' obtained using this method and the display sentence "red house" are written side by side on the original document 1400. Ta.

Figure 15 shows automatic translation between documents in different languages.

This is an example of expressing meaning using case grammar.

In case grammar, for the original sentence in Figure 15, ``I go to school by train at 8 o'clock,'' the verb ``go'' is the main character, and the action is ``I'' and 9 o'clock is ``8 o'clock.''

The means is "train" and the end point is "school".
The meaning of the entire sentence is expressed by clearly indicating the semantic roles of the words in the sentence. If this semantic expression is written as higher-level information in the form shown in Figure (c) and written alongside the original text for humans, then the machine-readable information can be used to generate a translated text as shown in Figure (d).
o to the 5choolat 8 o'clo
ck by trajn," can be automatically generated.

Information protection methods against contamination or loss of information media include:
Machine-readable information may be expressed and written in several different ways. For example, a method using a bar code for optical scanning and magnetic ink for magnetic reading, a method for turning a part of an information medium into microfilm, a combination of the knob method and the above method for embedding a semiconductor storage device in an information medium. etc.

Fig. 14 shows an example in which the present invention is applied to interlingual translation processing of a document containing both text and figures, but this is a simple example and one of the most important technological development issues in recent document processing. The application to the process of extracting specific information from a document will be explained using FIG. 16.

FIG. 16 1500 shows a text section 1504 and a graphic section 150.
It is a document that has both. A text start symbol based on the machine-readable code of the present invention is entered in the start column portion (1510, 1512) of the text portion, and the end column portion (1514,
1516, 1518), enter the text end symbol.

Further, a figure start symbol is written in the start column part (1520, 1522) and a figure end symbol is written in the end column part (1524, 1526) of the figure part. By scanning the document created as described above in the horizontal direction using, for example, the apparatus shown in the embodiment in FIG. It is possible to realize functions such as selectively extracting only the 1510 in the diagram
．． Instead of starting or ending symbols using machine-readable codes written in the 1514th copy, etc., it is possible to store information about text parts and graphic parts in a place different from the visible information, but compared to these methods, In the above embodiments, it is advantageous to use a method in which information for humans and information for machines can be overwritten.

(1) Document area can be used effectively.

(2) When printing information on a document, it is common for the printed information to be misaligned or rotated rather than being positioned as planned with respect to the original paper surface. If the information for humans and machines is written in separate parts, the durability against such disturbances will be weakened.

For example, the origin position of the text portion 15o4, which is displayed in a machine code with the upper left point of the page as the starting point, deviates from the actual origin position, causing an error when the document is read again. On the other hand, in an embodiment in which information for machines and information for humans are written overlappingly in the same place, both pieces of information are equally subjected to the same disturbance35), so errors when rereading are reduced. In other words, the ability to withstand external disturbances is strong.

〔Effect of the invention〕

According to the present invention, by creating a document including text and graphics using information that is visible to humans and information that is difficult for humans to see but is readable to machines, the following effects can be achieved.

(1) There is no longer a need for double management, as in the past, in which a paper document is created for humans and a floppy disk is used to store the machine code of the document for machines.

(2) It becomes possible to directly read a document, including its semantic information, by a machine. Furthermore, by multiplexing machine information and expressing it on a document, even if part of the document is damaged, the original document can be easily restored. The above means that it is now possible to directly decipher general documents by machine, which has been considered extremely difficult in the past. It is also possible to automatically read the circuit diagram ■I.

(3) Automatic translation of documents between different languages can be easily realized by writing semantic information of sentences and figures on documents in a machine-readable form. Previously this was not possible.

[Brief explanation of the drawing]

FIG. 1 is a conceptual diagram showing the overall outline of the present invention, FIG. 2 is a diagram showing one embodiment of the document representation method of the present invention, FIG. 3 is a diagram showing another embodiment of the document representation method of the present invention, and FIG. FIG. 4 is a diagram showing an example of encoding machine-readable information in the document expression of the present invention, FIG. 5 is an explanatory diagram of the input/output section of the document of the present invention, and FIG.
The figure is a block diagram of the input/output function of the document of the present invention, Figure 7 is a diagram showing an example of the document editing software of the present invention, and Figure 8 is a diagram of an embodiment of the high-order/low-order information conversion dictionary used in the present invention. , FIG. 9 is an explanatory diagram of an embodiment of the output processing main part in the present invention, FIGS. 10(a) and (b) are explanatory diagrams of a partial embodiment of the input processing main part in the invention, and FIG. Fig. 11 is an explanatory diagram of the relationship between high-order information and low-order information used in the present invention, Fig. 12 is an explanatory diagram of the processing contents of the terminal output processing unit in the present invention, and Fig. 13 is a diagram showing partial damage to which the present invention is applied. FIG. 14 is a diagram showing an embodiment of a document editing method using direct document input; FIG.
FIG. 5 is a diagram showing another embodiment of the automatic translation system between foreign language documents to which the present invention is applied, and FIG. 16 is a diagram showing an example of application of the present invention to document extraction processing.

Claims (1)

[Scope of Claims] 1. A document in which the display positions of both information in an information expression format that is visually visible to humans and information in an information expression format that is difficult for humans to see but readable to machines are overlapped. How information is expressed above. 2. The method for representing information according to claim 1, wherein a material containing magnetism or a material having characteristics of light reflection properties is used as the hard-to-see, machine-readable representation material. 3. The information expression method according to claim 1, wherein information that is difficult for humans to see visually but is readable by machines is expressed using an information expression substance that is visible to humans. A coding method called the MN code, which is typified by the 4, 2-7 code, is used to uniformize the appearance of the shading of characters, etc. on the document by changing the adhesion status of visible substances to the document. and machine-readable information. 5. A processing system comprising means for creating a document using the information representation method according to claim 1. 6. A processing system comprising means for reading a document created by the information representation method according to claim 1. 7. A computer system comprising means for simultaneously displaying visually visible information and difficult-to-see machine-readable information on a document, and means for reading the difficult-to-see machine-readable information. 8. A method for inputting and outputting the document according to claim 1 into a computer system using a dictionary in which pairs of visually visible information and data for generating difficult-to-visibly machine-readable information are defined for characters and figures. 9. The information representation method according to claim 1, wherein for each display of visually visible information, a plurality of pieces of corresponding difficult-to-visibly machine-readable information are displayed. 10. An information expression method having a function of selecting only one piece of information for the same piece of information from a plurality of pieces of machine-readable information expressed on a document according to claim 9. 11. A document editing device that reads, edits, or outputs an editing result of a document in which text and graphic information are displayed simultaneously using the expression method according to claim 1. 12. In a document using the expression method set forth in claim 1, semantic information of one or both of visually visible text and figures is retained as difficult-to-visibly machine-readable information, and by using this, it is possible to transfer between documents in different languages. A translation device that performs machine translation. 13. The translation device according to claim 12, wherein the term concept in predicate logic is used as the semantic information description means. 14. A processing system for controlling the operation of a computer system through document input using the semantic information description means according to claim 12. 15. The processing system according to claim 14, which is applied to the fields of medical consultation, legal consultation, resident list management, patent management, blueprint management, project management, painting management, music management, production management, library management, or education. 16. A document management system using a large-scale storage device in combination with the system of claim 6. 17. The computer system according to claim 7, further comprising means for recording and reading machine-readable information using a plurality of different recording means.