JP3204517B2 - Unknown word recognition method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to, for example, for the unknown word certification how in English and Japanese machine translation apparatus.

[0002]

2. Description of the Related Art In a conventional machine translation apparatus, a character string of an English sentence is input from a keyboard of an input device, character string data of a character code is formed, and a delimiter code such as a space, a comma, or a period in the character string is converted. It is broken down into lexicals as a material for recognition of lexical division, and a dictionary search is attempted for each character string of all lexicals. Then, the phrase whose phrase is present in the dictionary is determined as a registered phrase that is not an unknown word.

On the other hand, a phrase whose character string does not exist in the dictionary is determined to be an unknown word as an unregistered phrase, and a translator intervenes to determine whether the word is a typo, a variable name or an abbreviation symbol. The correction and other processing are performed.

[0004]

In technical documents such as computer manuals, a large number of words and phrases meaning command names, variable names, and the like appear. These lexical terms are words that appear frequently in specialists, but are not generally present and are not registered in dictionaries. Even if it is known that a phrase meaning a command name, a variable name, or the like is a word that does not exist in the dictionary, a dictionary search process for all the phrases is attempted, and the above-described determination process is followed.

On the other hand, in technical characters, general words are expressed in a normal typeface, that is, a standard typeface, but command names,
The lexical meanings of conversion names and the like are often expressed in special fonts other than ordinary fonts, that is, non-standard fonts such as bold bold font and italic font.

[0006] Therefore, if a dictionary expressed in a standard typeface is subjected to a dictionary search and a phrase expressed in a non-standard typeface is excluded from the dictionary search, the chance of dictionary search for unknown words is greatly reduced. Can be done.

The present invention has been made in view of the above points, and does not perform a dictionary search for an unknown word without removing a phrase expressed in a non-standard typeface, thereby greatly reducing the chance of a dictionary search for an unknown word. an object of the present invention is to provide an unknown language certification how that was.

[0008]

SUMMARY OF THE INVENTION The present invention provides a machine translation apparatus.
, For words determined to be unknown words in the original sentence ,
An unknown word certification method of instructing the intervention to the operator, original
Enter the document image is an image data, a character image cut out the image data to a character unit, is converted into text data of a character code to recognize text image data of the character image, and from the character image performs recognition processing or the standard typeface or non-standard typeface, the divided text data into words by a delimiter, structure the words with the character of a non-standard typeface
If the word is unknown,
A determination target outside the process as, when the word is composed of characters of a standard font that does not contain non-standard typeface characters,
Perform dictionary search of the word, unregistered word to the dictionary, it is determined that the unknown word, and controls so that the referent to the operator.

[0009]

According to the present invention, when performing a dictionary search for unknown word recognition, it is determined whether or not a search target lexical character is described in a standard typeface in a document.
Words in non-standard typeface descriptions are largely excluded from unknown word recognition, greatly reducing the chance of dictionary search for unknown words.

[0010]

FIG. 1 is a functional block diagram of an embodiment for carrying out the present invention.

In the character recognition device 13 shown in FIG. 1, reference numeral 1 denotes a document image input means for inputting a document for which unknown word recognition is performed into a computer as image data, such as a scanner. Reference numeral 2 denotes an image data storage unit for storing document image data read by the document image input unit 1, and 3 denotes a character image that is accurately cut out of the document image input by the document image input unit 1 in the order in which a sentence is composed. The character image recognition means 4 is a character recognition means for recognizing the character image cut out by the character image recognition means 3 as a character and outputting a corresponding character code, and 5 is a character typeface for determining the typeface of the character in the character image. It is a determining means.

Reference numeral 6 denotes a character code as a result of character recognition,
This is text data storage means that is stored together with typeface information.

In the lexical character recognition device 14, 7 is a lexical character recognition means for certifying a lexical character from a character string stored in the text data storage means 6 on the basis of a delimiter such as a period, a comma, a space, and the like. Lexical data storage means for storing the information of the lexical character recognized by the means 7 together with the information of the typeface.

In the dictionary device 15, reference numeral 10 denotes the lexical recognition means 7.
Is a dictionary for knowing whether or not the phrase recognized by is a unknown word, and 11 is a dictionary search means for searching the dictionary 10 using the phrase as a key. Reference numeral 12 denotes an overall control device that controls the devices 13, 14, and 15 and the text data storage means 6.

Hereinafter, the operation of FIG. 1 will be described with reference to the flowchart of FIG. 2 taking an example of translating English into Japanese.

Document Image Input Processing (S ₁ ) The English characters of the English document input by the document image input means (scanner) 1 are stored in the image data storage means 2 as image data.

Here, it is assumed that the original text in Table 1 is input to the character recognition device 13. Two a to see English fonts
rgv and one exec are bold bold fonts (underlined), but have a different typeface from other lexical standard fonts. The former typeface is called "non-standard typeface". The latter other typeface is referred to as a "standard typeface".

[0018]

[Table 1]

Character Recognition Processing (S ₂ ) In the character recognition device 13, the character image recognition means 3 separates and cuts out image data for each character from the image data in the image data storage means 2. The cut-out character image data is passed to the character recognition unit 4. The character recognition unit 4 performs a character recognition process on the image data, and stores the obtained character code in the text data storage unit 6 as character data.

Further, at the time of character recognition, the character type determining means 5 determines the type of the character image, and stores the type information in the text data storage means 6.

As a result, font information in the original text is set as attribute information in each text data. As a result of performing this character recognition processing on all the characters in the image data input in this way, the input document is converted into a text code string and stored in the text data storage unit 6. Will be.

As a result of performing the character recognition processing on the input sentence, the information in Table 2 is logically stored in the text data storage means 6.

[0023]

[Table 2]

Here, the typeface attribute column indicates a distinction between a standard typeface and a non-standard typeface. 0 is a standard typeface and 1 is a non-standard typeface. Typeface information exists corresponding to each character code.

Lexical qualification processing (S ₃ ) In the lexical qualification device 14, the lexical qualification means 7 performs lexical qualification processing on the text information stored in the text data storage means 6. Delimiters such as blanks, commas, and periods are lexical recognition materials. The result of the lexical recognition processing is stored in the lexical data storage means 8.
Performing the lexical recognition process on the above example, logically,
The results in Table 3 are stored in the lexical data storage means 8.

[0026]

[Table 3]

In the lexical recognition processing, the lexical style determining means 9 performs lexical style determination on the lexical recognition result in the lexical data storage means 8. Typeface information associated with each character code is used for this lexical typeface determination.

As a result of the lexical font determination, the lexical information in the lexical data storage means 8 is appended with the lexical unit information shown in Table 4. If the font information is 0, it is a standard font, and if it is 1, it is a non-standard font.

[0029]

[Table 4]

Unknown Word Recognition Processing (S ₄ ) FIG. 3 is a detailed flowchart of the unknown word recognition processing shown in FIG. The unknown word recognition is performed by performing a dictionary search by the dictionary search means 11 based on the lexical typeface. First, for each phrase in the phrase data storage means 8, the first one phrase is extracted (S5). If the typeface of the lexical character is a non-standard typeface (S6N), it is excluded from unknown word recognition. If the font is a standard typeface (S6Y), the dictionary 10 is searched, and the dictionary 10 is searched using the lexical key as a key.
(S7). As a result of the dictionary search (S8), the unregistered phrase is determined to be an unknown word, and information on the phrase is registered as an unknown word in the unknown word data management unit of the text data storage unit 6 (S9). ).

Next, when there is a remaining phrase in each phrase in the phrase data storage means 8 (S10), the process returns to (S5) of the loop and repeats the dictionary search. When there is no remaining lexical character in the lexical data storage means 8, the process exits from this loop and performs a translation word selection process (S11).

According to the present invention how, in the above sentences, will be the presence of an unknown word not observed. However, in the conventional unknown word recognition method, the result is as shown in Table 5. The underlined words in Table 5 are the words recognized as unknown words. These words mentioned are not unknown words that the user needs to correct.

[0033]

[Table 5]

Application Example Next, it is assumed that the original text is as shown in Table 6.

[0035]

[Table 6]

Here, the following misspelled words are intentionally included.

1. array (wrong) ... ar
ray (correct) 2. parmeter (wrong) ... paramete
r (correct) In the conventional method, the unknown word recognition result shown in Table 7 is obtained because the difference in typeface is not considered. Five words indicated by underlines are pointed out as unknown words. However, except for three non-standard typeface descriptions, the words to be corrected by the user are only the two words in error described above.

[0038]

[Table 7]

[0039] However, <br/> unknown word certified how according to the present invention, the lexical had been expressed in different typefaces in advance original sentence, since the processing in consideration of this fact is made, as shown in Table 8 It is possible to accurately point out only unknown words.

[0040]

[Table 8]

[0041]

Unknown word Certified <br/> how the present invention described above, according to the present invention, upon unknown word certification process, in a document of the original, the phrase phrase typeface has been represented by like italics , to become the outside of the object of unknown words certification process, it is possible to suppress the occurrence of the unknown word candidates. Accordingly command or variable name, such as the words of a non-standard typeface characterized only in a document without the unknown word candidate, intends rows dictionary searching only terms of a standard typeface as unknown word candidates it is possible to target <br/> precise indication of unknown words by, the translator, that Do that typographical other, necessary minimum of the unknown word candidate is presented
Therefore, unnecessary complications in the correction processing are reduced.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an embodiment for implementing the present invention.

FIG. 2 is a flowchart showing the operation of FIG.

FIG. 3 is a detailed flowchart of an unknown word recognition process in FIG. 2;

[Explanation of symbols]

1 ... document image input means, 2 ... image data storage means,
3 ... character image recognition means, 4 ... character recognition means, 5 ... character style determination means, 6 ... text data storage means, 7 ... lexology recognition means, 8 ... lexical data storage means, 9 ... lexical style determination means, 10 ... dictionary , 11: dictionary search means, 12: overall control device, 13: character recognition device, 14: lexical recognition device,
15… Dictionary device.

Claims (1)

(57) [Claims] An unknown word recognition method for instructing an operator to intervene in a machine translation device for a word determined as an unknown word in an original sentence, wherein a document image as an original sentence is inputted as image data. Extracting the image data in character units to form a character image, recognizing the character image data of the character image, converting the character image into text data of a character code, and recognizing whether the character image is a standard typeface or a non-standard typeface was carried out, the divided text data into words by delimiters, when the word is composed of non-standard typeface characters, the single
Was excluded from the determination process as unknown words without dictionary search word, when the word is composed of characters of a standard font that does not contain non-standard typeface character performs a dictionary retrieval of the word dictionary unknown word certification method unregistered word determines that unknown words, and controls so that the referent of the operator.