JPS60163171A - text reading device - Google Patents

Abstract

PURPOSE:To input sentences easily from newspapers, magazines, etc. by inputting an indicating code to read automatically a sentence, which is separated in paragraphing or physically, as one set. CONSTITUTION:The code indicating sequence is added to an optionally closed plane pattern to indicate a part, which should be read as one sentence, on the surface of a paper to be read. The form to which the indicating code is added is set to a reading part 22, and the closed pattern is read by a pattern reader 22-2, and blank parts are detected in the closed pattern in accordance with character arrangement (vertical writing or horizontal writing) to recognize paragraphs, and the pattern is fractionized to unit patterns by column or row, and characters are read in the order of pattern number by a character reader 22-1 and are stored in a file.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a device for reading characters, particularly sentences written in a natural language.

[Background of invention]

Conventionally, when reading text, the OCR (optical character reader) is used to recognize all characters character by character. It is only read in units,
Either they are not aware of where the sentences are connected, or even if they are aware of it, they only say ``When one line ends, it continues on to the next line,'' as shown in Figure 1 a. It was not possible to recognize staircase-like structures, those with columns as shown in the same figure, and even highly ambiguous characters (Q), such as C or d in the same figure, as a single group of trenches.

[Purpose of the invention]

An object of the present invention is to provide a device for easily inputting sentences by automatically or by inputting simple instructions to recognize sentences σ) connections having stepped means as one unit of a unified sentence. It is something.

[Summary of the invention]

In order to achieve this objective, the present invention provides a device that reads codes for indicating the range and order of paragraphs, etc., and uses this information to recognize paragraph breaks and order, and physically separate them. It provides a means to recognize a group of characters that may or may not exist as one connected sentence.

[Embodiments of the invention]

Hereinafter, the present invention will be explained in detail with reference to Examples. ,,
2 and 3 show an example of the configuration of the present invention,
Keyboard 21, reading section 22, (character reading device IIL2
2-1, figure reading device 22-2), processed iron [(
(hereinafter referred to as CPLI) 24, a main memory 25, a file 26, etc. The keyboard 21 is provided with a vertical writing/horizontal writing instruction key 31, a file name specification key 32, and a start key 33.

The main memory 25 includes a processing program 51, a figure position table 52, and a partial buffer memory 53, as shown in FIG.
, including the entire panzer memory 54.

Next, the processing operation of the present invention will be explained.

A person (hereinafter referred to as a user) who uses the text reading device of the present invention instructs which part of the text should be read as a whole. The method of instruction is to apply a code indicating the order to any closed plane figure (hereinafter, the figure indicating the range and the code indicating the order are collectively referred to as the instruction code), or to write a line between the closed figures. By tying with. One embodiment is shown in FIG. 1f.

When connecting with a line, according to the general connection rules for vertical/horizontal writing (for example, ``If it is written vertically, the center of gravity of each shape is closest to the top right, and if it is equidistant, the one on the right takes precedence''). It is necessary to use them within the range where the order is clear.

In the following, the method of connecting with lines will also be described as a method of assigning numbers equivalent to this method.

The code indicating the order may be numbers or letters such as "ABC" or "Iroha", but it is preferable to use an easy-to-read code indicating the order, and in order to make the bladder removal easier and more reliable, use pre-registered letters. It is also possible to give instructions using something like a mark.

The closed plane figure may be a free figure as shown in FIG. 1e. In that case, depending on the information of the vertical writing/horizontal writing instruction key 11, or the CPU 4
-2 gives the position of the figure ti! it: Full reading, character group inside is read by character reading device 22-1 or character nF
k Blur processing (characters are black dots, black dots are placed at certain positions)
For example, vertical writing/horizontal writing is recognized by a method of automatically determining by having the graphic reading device 122-2 read a graphic (processing in which a certain area around the graphic is determined to have a black dot). Based on this information, the figure reading device 22-
2 alone or in conjunction with the character reading device z2-1,
Determine the range of each row (column). This can be determined by setting the end point of the line at the position where the debunked relevance ends in the vicinity indicated by the figure. By doing this, even if a figure is touching a character during freehand input, the range of the character group can be recognized correctly, and you can input without being too nervous. , you can specify the range by freehand.

You cannot enter the order code only from the shape.
This is because the order of each part of a sentence represented by a plurality of figures may become unclear. For example, Figure 1C and d are
Without numbering, it is unclear which order is correct. Furthermore, in the case of vertical writing, the order 3.4 in Figure 1 can be understood by humans even without any ordering, but in an example like Figure 1 f, it becomes unclear.

It is desirable that the instruction code be written in an ink or color different from that of the text, but this is not the case as long as the graphic reading device 22-2 can distinguish it from the text. Alternatively, the instruction code may be written on a completely separate sheet of paper (for example, a transparent sheet such as 70) and read separately from the text. In this embodiment, the explanation will be made on the assumption that the instruction codes are written in different colors on the same sheet of paper, so that the instruction codes can be read independently of the text.

The process proceeds as follows. First, the paper to which the instruction code has been added can be held in the reading unit 22 (the part where the paper is placed is common to the character reading device 22-1 and the graphic reading device 22-2, so this will be collectively referred to as the reading part. 22) and the start key 33 is pressed, the CP
U4 starts all of the following processes according to the processing program 51.

First, the figure bladder reader @22-2 is activated and a closed figure is read. The method is not particularly defined in the present invention. There is no particular restriction on how to represent a read closed figure, but in this example, if the figure is approximated by a polygon consisting of only vertical and horizontal lines, and the figure is expressed only by its vertices. Let's talk about. FIG. 6(a) shows an example in which a figure as shown in FIG. 5(a) is displayed in the figure position table 52. The x, ysx column 523 and Y column 524 in Fig. 6 are the horizontal axis and the vertical axis, respectively. - For example, in the case of vertical writing, the origin is set at the upper right corner, and the axis extends to the left and bottom. It shows the position coordinates of the case.

The figure reading device 22-2 further reads the order code and stores it in the figure number column as illustrated in FIG. If the ordinal code is not folded, e.g.
If it is written vertically, numbers are assigned in order such as ``the center of gravity of each figure is closest to the top right, and if the distance is the same, the one on the right takes priority.'' For example, if there is no order code in the case shown in FIG. 1C, the order of numbers 2 and 3 from the top right is closest, and if they are the same, the right one takes priority, resulting in the order shown in FIG. 1D. In addition, for the shell point of each intervening figure, there is a spot point column 522.
Vertex numbers are assigned in order. - As an example, an example in which numbers are assigned counterclockwise from the upper right corner is shown in the X column 523 and Y column 524 in FIG. In the following, to simplify the explanation, sentences written vertically will be described unless otherwise specified.

Further, the following description will be made assuming that the order is numbered.

Once the vertices of each closed figure and the processing order between closed figures are determined, the process is first proceeded for each closed figure, and finally they are grouped together.

A flowchart of one embodiment of the above processing is shown in FIG.

This will be explained below with reference to FIG.

First, the maximum value of the figure number and i vc □ are stored in MAX (step 101). Next, it is checked whether i is larger than MAX (step 102). If not large, l
is increased by one (step 103), and the third figure is divided into single shapes (step 104).

This is a figure that cannot be further divided into a single form,
Figure 5 CV Refers to a step-like structure as shown here, with no paragraphs inside. Items that contain internal paragraphs, such as those shown in FIG. 5, are not single-occurrence forms, and these are separated into single-occurrence forms in step 104 (an example of this method will be described below). Next, start the character reading device 22-1 for each single form,
The characters are read and stored in the partial buffer 53 (step 105).

When all the characters for each figure have been read, the entire character is transferred to the entire buffer 54, and the partial buffer 53 is stored for the next reading (step 106). Sequentially, all figures are read, and when l becomes MAX or more, r[ is assumed to be complete, and input in advance, or after completion of input,
By instructing the user to specify the file name, the file name input using the file name specification key 32 is added, and file 2 is input.
6, and the entire buffer is nitrided to complete the process (step 107).

Among the above processes, step 104 and step 105
This will be explained in further detail.

A flowchart of an embodiment of the process of step 104 is shown in FIG.
As shown in the figure. This processing is for dividing, for example, a figure including a paragraph as shown in Figure 5 into single figures.

Instead, look at the shape in Figure 1 or from below. In Figure 8, from above. The spacing between the following rows can be arbitrary as long as it does not skip over paragraphs.

The process starts from the topmost row of the housing, and a reading position is set (step 201).Next, this row is read aloud (step 202), and it is checked whether all the rows are blank (step 203). In this embodiment, the space between paragraphs is described as a blank space, but a line or the like indicating a break may also be used. In that case, by reading the characters, we can see that there are other characters σ〕 and q〕, so the code other than those characters is blank and 1.
By regarding it as m4, etc., the same processing as in this embodiment is possible. If there is anything other than 9 white, proceed to the next row (step 206). If all of the rows are blank, it is determined whether the blanks are wider than the character spacing by a clear VC (step 204). If it's not wide, there's an accidental space between letters'k
It is assumed that the row is U and moves to the next row (step 206).
). If it is determined that it is clearly wider than the character spacing, it is assumed that there is a paragraph and the graphic is divided into vertical sections (step 205), and the process proceeds to the next row (step 206). In step 206, it is checked whether all horizontal rows (that is, the vertical axis) are finished, and if so, the process ends. For example, if there is a figure with paragraphs (sets) as shown in Figure 5, and the position of this figure is as shown in Figure 6a, two paragraphs will be found by the above process, and as shown in Figure 8A5, d. It can be divided into three single figures. An example of displaying the position of each figure is shown in Figure 6 b.
'(show.

Next, the sting 105 in FIG. 7 will be revised and described.

This process does not have a step I@ as in FIG. 5C, but a stepped V? This is the process of reading the characters in the figure that has become -.

In this case, the processing may be performed by recognizing the range of each column VC from the graphic position. In that case, if the column is X
If the position is =3, this if line and figure position table 5
By finding the intersection with the figure displayed in 2, 5≦Y
≦15 is found in each station. ), the figure may be divided vertically from the vertex into rectangles, and then the range may be read all at once.

〔Effect of the invention〕

As explained above, according to the present invention, texts that are physically separated in columns can be automatically read as a single unit by inputting all the simple instruction codes, so text from newspapers, magazines, etc. You can easily enter text.

[Brief explanation of the drawing]

FIG. 1 is a plan view showing some examples of text columns according to the present invention, FIG. 2 is an external view of an apparatus according to an embodiment of the present invention, and FIG. FIG. 4 is a diagram showing an example of the internal configuration of the device; FIG. 4 is a diagram showing an example of main memory allocation; FIG. 5 is a diagram showing an example of a stepped character string figure for explaining the present invention; FIG. 6 is a diagram showing an example of a figure position table, and FIGS. 7 and 8 are flowcharts showing an example of the reading processing procedure in the apparatus IIC of the present invention. ¥J 1 Figure ((L) (1)) (C) ((t) ■ Z Figure 2226 χ 3 Figure 5 ■ 4 Figure 'A+Figure (β-) <b) Tomi 7rj! U

Claims (1)

[Scope of Claims] 1. A character reading device that reads characters, a graphic reading device that reads limited codes such as simple figures and numbers, and a control device that controls these character reading devices and graphic reading devices. What is claimed is: 1. A character reading device comprising: and a file for storing the reading results, and reads characters according to information obtained by the graphic reading device. 2. Claim 1, which is characterized in that the range to be read is indicated by a handwritten closed figure, and this t'L is read by a figure reading device alone or in conjunction with a character reading device.
Text reading device described in section. 3. A patent claim characterized in that a visceral ordinal coat is added to mutually independent closed figures, and by reading all of these, even if the sentences are separated, the order can be known and the sentences can be read as a single sentence. A text reading device according to scope 1. 4. A patent characterized in that mutually independent closed figures are connected by their legs, and their order is known by this and general connection rules for vertical/horizontal writing, so that even if the sentences are separated, they can be read as one sentence. A text reading device according to claim 1. 5. The text reading device according to claim 1, wherein when fully determining whether a text is written vertically or horizontally, all line directions are recognized by blurring processing. 6. Even if a handwritten closed figure is not drawn accurately, it is possible to specify a fairly free range by determining the cut point between the 9 spaces in the vicinity. A text reading device according to claim 1.