JPS58222330A - Japanese language converting system - Google Patents

Abstract

PURPOSE:To select quickly and efficiently Kanji (Chinese characters), by adopting a two-stage selecting method, in a Japanese language converting system converting an inputted word into a Kanji. CONSTITUTION:The result of recognition transmitted from an input terminal 10 is stored in a character matrix buffer 1 and the word part to be converted into a Kanji is stubbed. The character matrix corresponding to the stubbed word is given to a Kanji converting section 2, where it is converted into a Kanji code. Further, a part other than the word is given to a Kana (Japanese syllabary) character converting section 3, where the part is converted into a Hiragana (round form of Japanese syllabary) or Katakana (square form of Japanese syllabary). Finally, a Kana/Kanji mixed sentence forming section 4 forms a Kana/Kanji mixed sentence by combining outputs of the Kanji converting section 2 and the Kana character converting section 3.

Description

[Detailed Description of the Invention] This invention provides a method for quickly selecting a desired kanji from a large number of kanji obtained as a result of comparing the recognition result with a word dictionary when inputting a Japanese sentence using voice recognition or character recognition technology. This article concerns a Japanese conversion method that has a selection function.

Background> As Japanese information processing progresses, monolingual input using voice and handwritten characters is seen as a promising technology as a simple Japanese input method. In the case of voice input, input is performed in monosyllable units, and each syllable is first recognized by the voice recognition unit, and then
A process of comparing the words with a word dictionary and converting them into kanji is performed. In this case, the following processing can be considered for general K. That is, since current speech recognition technology cannot always completely recognize each syllable, it is common to output several candidates for each syllable. An example of such a recognition result is shown in FIG. This is the speech recognition result when "hot spring" is input, and the it# clause a1 is recognized as either "ko", "o" or "ho", and the second syllable al is recognized as "n", 3rd syllable a
This is a case where s is recognized as either "se" or "ze", and the fourth syllable a4 is recognized as "n".

Since this format is not necessarily used for voice input, but is also used for handwritten kana character input, it will be referred to as a character matrix from now on. To convert this character matrix into kanji, select one character from the multiple candidates for each character, match the resulting string of characters with the word dictionary, and if a matching heading exists in the word dictionary, Convert to kanji corresponding to the heading. Since there are homonyms in Japanese, there is always only one type of kanji that corresponds to one heading.For example, the kanji that corresponds to the heading ``ondo'' include ``temperature'' and ``ondo.'' However, when the input is a character matrix, a plurality of kanji characters are obtained as a result of dictionary matching for each different character sequence generated by the combination K of candidates. For example, in the example in Figure 1, the # of the kanji ``hot spring'' corresponding to the character series ``onsen'' is
The kanji character ``cross line'' for the one or two character series ``konsen''.

For the character series ``konsen'', the kanji ``before marriage'', ``chaosen'', etc. are obtained. Therefore, the number of kanji that can be obtained from one character matrix is extremely large.・□・1: In the conventional kana-kanji conversion method using kana keyboard input, the input is a series of characters, so even if there are homophones, the number of kanji is limited to a small number. Therefore, the usual method was to display kanji characters one after another on the screen as homonyms were used by the user on the keyboard, and to have the user select the correct kanji character from among the kanji characters. On the other hand, when the input is a character matrix, there are a large number of kanji candidates as mentioned above, so
The method of displaying these on the screen one after another as the user operates the keyboard and letting the user make a selection is not only extremely complicated and takes 9 hours to operate, but also places a heavy burden on the user. This means that an efficient Japanese input device cannot be realized.

<Summary of the Invention> An object of the present invention is to provide a Japanese conversion method that can quickly select Kanji characters from an input character matrix and is easy to operate.

In this invention, the user is first asked to select which reading the input word belongs to, and then the user is made to select the correct kanji from a plurality of kanji that belong to the selected reading ``:''. By using this selection method, the correct kanji can be efficiently selected from among a plurality of kanji.

<Embodiment> FIG. 2 shows an embodiment of this invention, in which character matrix buffer 1
has the function of holding the recognition results sent from the speech recognition unit (or handwritten character recognition unit), which is not shown in this figure, through the input terminal 1ot-, and cutting out the part of the word to be converted into kanji. ing. The character matrix corresponding to the word cut out by the character matrix buffer 1 is sent to the kana-kanji converter 2 and converted into a kanji code. Also, parts other than words are sent to the kana character conversion section 3. The kana character conversion unit 3 performs processing such as converting hirakana characters into hiragana codes and katakana characters into katakana codes. Finally, in the kana-kanji mixed υ sentence creation section 4, the kanji conversion section 2.
The outputs of the kana character conversion section 3 are combined to create a kana-kanji mixed shi sentence.

Kana-Kanji Conversion Section> FIG. 3 shows in detail the configuration of the Kana-Kanji conversion section 2 in FIG. 2. The character matrix to be converted into kanji is sent to the matching unit 2.
Sent to 1. The word dictionary 22 stores kana headings expressed as character sequences and their corresponding kanji codes.

The matching unit 21 matches the character matrix with the word dictionary 22.

That is, select one candidate for each character in the character matrix, and check the word dictionary 22 with the character sequence formed by connecting these candidates.
If a matching kana heading exists, the corresponding kanji code is output to the kanji candidate buffer 23. The matching unit 21 performs this process for all character sequences that can arise from the character matrix. As a result, kanji corresponding to a plurality of readings are found in the kanji buffer 23, and a plurality of kanji, so-called homonyms, are stored for each reading.

The control section 24 allows the user to select the correct Kanji from among these many Kanji characters using the display section 25 and the keyboard section 26 through the following process. The first step is to have the child choose the correct reading. There are several possible methods for this selection. For example, all the different pronunciations may be displayed on the display section 25, and the user may be asked to type a number or the like corresponding to the correct pronunciation from the keyboard section 26. It is also possible to provide a key for instructing the display of different readings, and each time the user presses the corresponding key, a different reading or the corresponding kanji are displayed one after another. The point is that the purpose of this invention is to have the correct reading selected in the first step.
The detailed method is not stipulated.

Once the correct reading is selected through the above process, the next correct kanji is selected. In this case as well, several selection methods can be considered. For example, all the kanji characters corresponding to the corresponding pronunciation may be displayed on the display section 25, and the user may be asked to type a number or the like corresponding to the correct kanji character from the keyboard section 26. Alternatively, a key for instructing the display of homophones may be provided, and the homonyms may be displayed one after another each time the user presses the corresponding key. The point is that the purpose of this invention is to select the correct kanji from the homophones in the second step, and the detailed method is not specified.

When the correct kanji is selected through the above processing, the code of the corresponding kanji is sent to the kana-kanji mixed script creation section 4.

By adopting the above-mentioned two-stage selection method, it becomes possible to efficiently select the correct kanji from among a large number of kanji candidates. For example, if there are M types of different pronunciations corresponding to a character matrix and N kanji corresponding to each pronunciation, a method of displaying kanji candidates one after another on the screen by keystrokes on the keyboard can be used up to MXN times. Although key operations are required, the method of the invention allows the correct kanji to be selected with a maximum of M+N key operations, making it possible to perform high-speed Japanese conversion.

In addition, in the example mentioned earlier, from the input character matrix,
Then, we are checking the word dictionary with the character sequence obtained by 9, so
Since each character series corresponds to a different reading of the character matrix, sets of kanji that belong to the same reading trap are automatically obtained. Before checking the word dictionary, each character sequence obtained is displayed in a character matrix and the user selects it.The selected character sequence is checked against the word dictionary and selected from among the homophones. You may also choose the correct kanji.

<effect> As explained above, this invention is a voice recognition technology.

It provides a Japanese conversion method for Japanese input devices using character recognition technology, and its gist is to select the correct kanji from a large number of kanji candidates obtained by comparing the character matrix of the recognition result with a word dictionary. In this case, a large number of kanji candidates are divided into multiple groups with the same reading and Kt, and the user first selects the correct reading, or the user selects the correct reading without dividing the kanji candidates into multiple groups, and then The purpose of this method is to use a second-level selection method in which the correct kanji is selected from homophones belonging to the same reading. As a result, it is possible to select the correct kanji from a large number of kanji candidates in a short time with less burden on the user, and this has the advantage of realizing a Japanese input device with excellent operability.

[Brief explanation of drawings]

Fig. 1 is a diagram showing an example of a character matrix, Fig. WJ2 is a block diagram showing an embodiment of the present invention, and Fig. 3 is a block diagram showing a detailed example of the kana-kanji conversion unit 2 shown in Fig. 2. . l...Character matrix buffer, 2...Kana-Kanji conversion unit,
3... Kana character conversion section, 4... Kana-kanji mixed μshi sentence creation section, 21... Collation section, 22... Word dictionary, 23.
...Kanji candidate buffer, 24...Control unit, 25...
Display section, 26...keyboard section. Patent Applicant Nippon Telegraph and Telephone Public Corporation Agent Takashi Single Needle

Claims (1)

[Claims] In the Japanese conversion method, an input word expressed as a character sequence (hereinafter referred to as a character matrix) with multiple candidates for each character is compared with a word dictionary to convert the input word into kanji. On the other hand, an operating means for selecting which reading the input word belongs to, and an operating means for selecting the correct kanji from a plurality of kanji belonging to the selected reading, are used to select the correct kanji from a large number of kanji. Japanese conversion method characterized by selection.