JPS5854431B2 - Misreading detection method for character recognition device - Google Patents

Description

[Detailed description of the invention] The present invention relates to a misreading detection method for a character recognition device.

Conventionally, many character recognition devices that read handwritten or printed characters have been announced, and recently, character recognition devices that read handwritten katakana characters have also appeared.

However, if the scope is expanded to include katakana, it is extremely difficult to completely eliminate misreadings, and a considerable number of misreadings are unavoidable at present.

From the user's point of view, it goes without saying that such misreadings should never occur, but if this is not possible, it is possible to detect any possible misreading, even if it is just a part of it. There is a strong desire to do so.

Therefore, an object of the present invention is to detect the possibility of misreading, especially in handwritten katakana recognition.

In order to achieve the above object, the present invention focuses on dakuten and handakuten.

In the handwritten katakana recognition device considered here, it is assumed that each character is written in an entry frame, and the voiced and handakuten characters are each written independently in one frame, as shown in FIG.

This entry method is used in the Telegram Bunraishin Paper, and is also widely used in ordinary katakana recognition methods.

In this way of writing, for example, the voiced and semi-voiced sounds are “Ga” → “Power” “S” "pu" → "fu" ".

'', so each clear sound part is recognized independently of the dakuten and handakuten.

So for example “Power” −→ “Ya” “Fu” → “Ko” Misreadings such as this may occur.

In the present invention, the misreading is detected by focusing on the fact that when such misreading occurs, combinations such as "ya" and "ko0" that are not allowed in Japanese occur.

Combinations of the above that are not allowed are: character that precedes Aye onani nune no mamim memo yayuyo Laril Leron (Uchitsuwa Kiwo) semi-dakuten (.

) preceded by the letters Naninunenomamimu memo yayuyo Laliru Lelowakion (Kakikukeko) This is the case.

Note that the letters in parentheses above indicate combinations that may be used in some cases; for example, ``力°'' has been made into a nasal sound, but it is occasionally used to express a leaky sound.

The present invention will be explained below with reference to Examples.

FIG. 2 is a block diagram of an embodiment according to the present invention.

The character recognition device 1 reads the characters on the form and sends a code representing the reading result to the inspection device 2.

In the inspection device 2, the determination circuit 3 determines whether the reading result is a voiced or half-voiced mark, and if the result is other than a voiced mark, it is interpreted as a correct reading, and the correct reading is indicated to the character recognition device 1. return signal.

After the above judgment is completed, the read result is stored in the temporary storage 4.

When the reading result is a voiced or half-voiced mark, the previous reading result is retrieved from the temporary memory 4.

The inspection device 2 sequentially compares this reading result with the code group stored in the ROM 5.

The code groups stored in ROM 5 represent kana characters that can precede dakuten or handakuten (or kana characters that cannot precede dakuten), and there is a difference in ROM 5 between those corresponding to dakuten and those corresponding to handakuten. is stored in the .

As a result of the above comparison, it can be determined whether the combination of the voiced mark or handakuten and the character preceding it is permitted.

If the combination is allowed, a signal indicating correct reading is returned to the character recognition device 1.

As a result of the above comparison, if the combination is not allowed, a signal indicating misreading is returned to the character recognition device 1.

When misreading is detected based on the signal returned from the inspection device 2, the character recognition device 1 displays the currently read character and the character immediately before it, and issues a warning to the operator.

Normally, correction characters are entered accordingly.

In the above explanation, we will use dakuten (or handakuten).

Although we assumed that the characters marked with (the same applies hereafter) are recognized before the voiced marks, the line in which the characters are entered may be scanned later depending on the observation device, and it goes without saying that the same method can be used in this case as well.

Also, in the above explanation, a warning is generated in the character recognition device 1.

Although correction and output are performed, the inspection device 2 may be provided with these functions.

As is clear from the above description, according to the present invention, misread characters can be detected with a simple configuration, thereby increasing the reliability of the character recognition device.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a procedure for writing handwritten characters, and FIG. 2 is a configuration diagram of an embodiment according to the present invention.

Claims (1)

[Scope of Claims] 1. A character recognition device for reading characters including a specific symbol, a memory means for storing types of characters for which combinations with the specific symbol are permitted or prohibited, and a character recognition device for reading characters including a specific symbol; determination means for determining whether or not a combination of the current and previous recognition results of is not allowed based on the content read from the memory means, and determining that the combination is an erroneous reading when it is determined that the combination is not allowed. A misreading detection method for a character recognition device characterized by: 2. The determining means includes means for determining whether or not the current reading result by the character recognition device corresponds to a specific symbol, and when the reading result corresponds to a specific symbol, The method is characterized by comprising means for determining whether or not the reading result of the current reading is to be combined with the reading result of the current reading based on the contents of the memory means, and determining that the combination is not allowed, determining that the reading is an erroneous reading. The misreading detection method according to claim 1. 3. The misreading detection method according to claim 1 or 2, wherein the specific symbol is a voiced mark or a handakuten mark.