JPH0378667B2 - - Google Patents

Description

[Detailed description of the invention]

(Industrial Application Field) The present invention relates to a search method for electronic dictionaries attached to various language processing devices including word processors, typewriters, and the like. (Technical Background) A spell check function is important in electronic word dictionaries attached to Roman word processors, Roman typewriters, etc., and there is a demand for such spell check processing to be performed as quickly as possible. Therefore, when performing a spell check,
For example, as illustrated in Table 1, it is conceivable to two-dimensionally group words to be registered in a dictionary by initial letter and number of characters, and to perform a search only within the group to which the corresponding word belongs. In addition, the first
The numbers shown in the table are the number of words belonging to each group. (Problem to be Solved by the Invention) However, according to the above search method, although the words registered in the electronic dictionary are divided into groups, the number of words belonging to each group is quite large. (For example, if the initial letter is “c”
(8-letter words are 1154 words), and the problem is that it takes a long time to search for the corresponding word. Therefore, it is difficult to use it for real-time processing on a typewriter or the like. In addition, in the above search method, each word is coded character by character and registered in the electronic dictionary, so the memory capacity required for each word varies depending on the number of characters. In the case of a large number of words, there is a problem that the storage capacity per word becomes large. Moreover, for words with a large number of characters, the search time per word will be quite long. Furthermore, according to the above search method, if a misspelled word is entered, it can only be determined as a misspelling after matching all the words in the corresponding group, which takes a long time. There is also the issue of becoming. (Means for Solving the Problems) The present invention aims to solve the various problems mentioned above. Therefore, in the electronic dictionary search method according to the present invention, when a plurality of words are stored in an electronic dictionary consisting of a memory device and searched, character weights are determined in advance for each alphanumeric character, and each position weight within the word is determined in advance. is determined, and the value obtained by multiplying the character weight and position weight for the first character of each word is divided by a predetermined prime number to obtain the remainder, and then for each subsequent character, the product value of the character weight and position weight is multiplied by the previous value. The operation of calculating the remainder by dividing the value obtained by adding the remainder by the above-mentioned predetermined prime number is repeated for each character up to the last character of the word, and the hash value obtained as the final remainder for the last character is used as an index of a certain number of upper digits. The data for each word is divided into data of a certain number of lower digits and registered in the above-mentioned electronic dictionary for each group with a common index, the above-mentioned hash value is calculated for the word to be searched, and the data of the search word and the above are registered. The method is characterized in that the index of the search word in the electronic dictionary is matched or mismatched with a group of data corresponding to the index. In that case, it is preferable to register the data groups belonging to each group in order of decreasing numerical value. (Example) Hereinafter, the present invention will be described based on an example with reference to the accompanying drawings and accompanying diagrams. The drawings show the general configuration of a language processing device to which the method of the present invention can be applied. In the figure, reference numeral 1 denotes an input device for inputting character and word information to the language processing device, and specifically, for example, a keyboard device,
Doublet device, OCR (optical character reader),
A magnetic tape device or the like is used. Reference numeral 2 denotes a storage device which is connected to the input device 1 and stores character and word information input and edited by the input device 1, and is, for example, a core memory, an IC memory, a magnetic disk device, or the like. Reference numeral 3 denotes an output device connected to the storage device 2 and outputting information stored in the storage device 2, such as various printers, display devices, magnetic tape devices, magnetic disk devices, etc. Reference numeral 4 is an electronic dictionary for spell checking connected to the storage device 2, and is composed of, for example, a core memory, an IC memory, a ROM (random access memory), a magnetic disk device, and the like. As will be described later, the dictionary 4 is equipped with an arithmetic processing unit dedicated to spell check processing, and can provide information as to whether the spelling of the input word is correct or not, depending on the timing from the storage device 2. ing. Further, 5 is a control device connected to each device 1 to 4,
For example, each device 1 is configured by a computer.
Controls the transmission and reception of signals between 4 and 4. Next, encoding of European words using the hashing method in the present invention will be described. In this encoding, first, a unique character weight (expressed in decimal notation for convenience) consisting of a binary number is determined for each character, as shown in Table 3 for some of the capital letters. Although omitted in Table 3, lowercase letters,
Character weights are similarly determined for numbers and the like. At the same time, as shown in Table 4, a unique position weight (expressed in hexadecimal for convenience) consisting of a binary number is determined for each character position within a word. Table 4 shows position weights for the first to seventh characters, but similar position weights are determined for the eighth and subsequent characters. Next, the hash value of each word is calculated in the following procedure based on the character weight and position weight. () In other words, first the first character of each word (the beginning of the word)
Multiply the character weight and position weight of . For example, in the case of the word "AIR", the character weight "60" of "A" is multiplied by the position weight "00080000" of the first character. In that case, it is preferable to divide the character weight into 3-bit units and multiply them by the position weight. Then, divide the multiplied value by the prime number closest to 2 ²⁷ to find the remainder,
Store the remainder. () Next, add the remainder obtained for the first character to the value obtained by multiplying the character weight and position weight of the second character, and divide that value by the above prime number to calculate a new remainder. Thereafter, the same operation is repeated for each character up to the last (end) character, and the finally obtained remainder is used as the hash value of the word. Here, the prime number closest to 2 ²⁷ is selected as the divisor for each division, so
All of the above hatch values are expressed as binary numbers of up to 27 digits. Note that by changing the divisor, the number of digits of the hash value can be changed arbitrarily. Table 5 shows the results of calculating the hash values using the above procedure for the beginning of the alphanumeric characters. Table 6 shows the first part of the hatch values obtained in this way, rearranged in ascending order (in ascending order of numerical value). In Tables 5 and 6, the hash values are expressed in octal for convenience. Once the encoding using the hash method is completed as described above, the hash value of each word is divided into the upper 11 digits (in the case of binary representation) of the index part (hereinafter simply referred to as index) and the lower 16 digits of the data part ( (hereinafter simply referred to as data), and grouped by words with common indexes. For example, among the word groups shown in Table 6, the hash value is
2 ¹⁶ The 43 word groups nevus~accommodator below (200000 in octal notation) are indexed “0”
It is classified as the first group. Further, a word group with a hash value of ²¹⁶ or more and less than ²¹⁷ is classified into the second group with the index "1". In this way, all words are classified into 2048 (=2 ¹¹ ) groups corresponding to the number of digits in the index. Note that the number of groups can be increased or decreased arbitrarily by changing the number of digits of the index. Table 2 shows the number of words belonging to each group (expressed in hexadecimal for convenience) when the number of groups is 2048. The vertical axis outside the margin of Table 2 shows the top three digits of the group number (each digit is expressed in hexadecimal), and the horizontal axis outside the margin shows the lowest one digit of the group number (expressed in octal). It is shown. As is clear from the table, by increasing the number of groups in this method, the number of words belonging to each group is reduced. By the way, as shown underlined in Table 2, in this method, the maximum number of words in one group is 88 (58 in 16 display), so all spell checking can be done with a maximum of 88 searches. It turns out. The word data grouped as described above are stored in the dictionary 4 in ascending order for each group. Furthermore, the start address of each group is determined and stored based on the number of words in each group shown in Table 2. (See Table 7). These addresses are used to select the relevant group during spell check. The spell check process using the dictionary 4 will be described below. When a word to be spell-checked (hereinafter referred to as a search word) is sent from the storage device 2 to the dictionary 4,
A calculation processing section (not shown) in the dictionary 4 calculates the hash value of the search word using the same calculation method as described above. Subsequently, after the relevant group is determined based on the index of the search word, the match or mismatch between the data of the search word and the data group belonging to the relevant group is sequentially verified. As a result of the comparison, if the data of the search word matches any data of the corresponding group, a signal indicating that the spelling is correct is transmitted to the storage device 2. On the other hand, if the search word data does not match any data in the corresponding group, a signal indicating that the search word is misspelled is transmitted to the storage device 2. In this method, the data in each group is arranged in ascending order, so especially in the case of a misspelling, the data for the search word will be smaller than any of the data in the data group of the corresponding group, and it will be the same as any previous data. If you do not, you can judge it as a misspelling at that point. By the way, as illustrated in Table 8, synonyms (synonyms) occur in this method. Here, a synonym means that two or more words have the same hash value. However, such synonyms are included in the total number of 72,000 words stored in Dictionary 4.
Since there are only 32 words and misrecognition occurs extremely rarely (1 word out of 2250 words), there is no problem in practical use. (Effects of the Invention) As is clear from the above explanation, according to the present invention, when grouping words and registering them in a dictionary, the number of digits of the index of the hash value is increased or decreased. The number of groups, in other words, the number of words per group, can be adjusted. In this case, by setting the number of digits of the index sufficiently large, the number of groups can be increased compared to the case of grouping based on the initial letter and the number of characters, and the number of words per group can be reduced accordingly. Therefore, the time required for spell checking can be shortened. In addition, in the grouping based on the initial letter and the number of characters, as long as the total number of words registered in the dictionary is constant, it is impossible to adjust the number of groups, that is, the number of words per group as in the present invention. Also, as previously mentioned, whereas conventionally each word was coded character by character and registered in the dictionary, the present invention uses the hashing method to code and register each word. The storage capacity per word can be kept constant and the storage capacity per word can be made sufficiently small. Therefore, the total dictionary capacity is also reduced. Furthermore, by arranging the data groups belonging to each group in ascending order, if a misspelled word is entered, there is no need to check the data of the input word with all the data of the corresponding group in the dictionary. Since a misspelling can be determined when the data of the input word becomes smaller than the data of any of the corresponding groups, the processing time can be shortened.

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】

【table】 [Brief explanation of drawings]

The drawing is a block diagram showing the general configuration of a language processing device to which the method of the present invention can be applied. 4...Dictionary.

Claims (1)

[Claims] 1. When storing a plurality of words in an electronic dictionary consisting of a memory device and performing a search, character weights are determined in advance for each alphanumeric character, and position weights for each position within the word are determined, For the first character of the word, the value obtained by multiplying the character weight and position weight is divided by a predetermined prime number to obtain the remainder, and then for each subsequent character, the value obtained by adding the previous remainder to the multiplication value of the character weight and position weight is calculated. The above operation of dividing by the predetermined prime number to find the remainder is repeated for each character up to the last character of the word, and the hash value obtained as the final remainder for the last character is used as the index of a certain number of upper digits and the data of a certain number of lower digits. The data for each word is registered in the electronic dictionary for each group with a common index, the hash value is calculated for the word to be searched, and the data for the search word and the search word in the electronic dictionary are calculated. A search method for an electronic dictionary, characterized in that a match or mismatch is checked with a group of data corresponding to an index. 2. The method according to claim 1, in which data groups belonging to each group are registered in order of decreasing numerical value.