JP2856775B2 - Document creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial application field) The present invention relates to a document creation device that converts a character string reading information into a desired character string and creates a document.
It relates to the function of preferentially outputting an appropriate one of the homophones output from the conversion as a result of the conversion.

(Prior Art) A document creation device converts reading information (a kana character string, a Roman character string input, a voice input, etc. from a keyboard) into a desired character string (a mixture of kanji, katakana, hiragana, etc.) and outputs a desired character. There is a function to get a line. At the time of this conversion, there are a plurality of words (same words) having the same reading in certain reading information, and there are also a plurality of possibilities at the cut portions of words / phrases.

Example 1: [Kinu-Function / Yesterday / Induction] [Crows-Wither / Wither / Crow / Crow] Example 2: [What is Hi (fire)-One person] Example 3: [Today (Today) the doctor (doctor) - today (today) to the dentist (dentist)] Accordingly, as a conversion result, the desired one can not necessarily be output from the beginning, a method of outputting the temporarily conversion result as a change possible candidates Is adopted. The document creation device has a function of selecting a desired word from other candidates for a word (example 1) having a homophone or a phrase (example 2) having a plurality of word breaks. When there are a plurality of segment breaks in the solid input reading character string (Example 3)
, A function of changing a segment break location and initiating re-conversion (phrase segmentation function) is provided.

However, if the conversion result candidates are displayed indefinitely, an excessive burden is imposed on the operator for the change operation, which is not preferable. Therefore, a measure is taken so that the most suitable conversion candidate is output first.

For example, it checks the grammatical connection between the preceding and following words / phrases to eliminate inappropriate conversion candidates, monitors the frequency of selection of homophones, and outputs the most frequently used words preferentially (in this case, , The field to which the document being created belongs (medical / legal / patent, etc.) may be taken into account.) The same phoneme selected before is likely to be selected next, so the selected word And output the word stored at the next conversion of the same reading in preference to other homophones.Remember the word pairs that are likely to appear in pairs and prioritize that pair over other homophones. (For example, register a pair of train-riding (Ru) so that a conversion result such as "Ride immediately to a reporter" is not output for "Get on quickly" ), Perform semantic analysis of conversion results and eliminate semantically ineligible conversion candidates That (for example, if the animal is subject "rather than" is correct, if human-subject meaning analysis such as "cry" is correct) there is,.

In the above method, there is a possibility that a plurality of correct cutting points exist, and there is a possibility that a plurality of homophone words exist in one cutting candidate, and an appropriate conversion result cannot be obtained unless used in combination with another method. In the method of, there is a case in which the conversion result is pulled too much by a frequently-used word or a previously selected word, and other homonyms are hidden, and the result is worsened. For example, in the conversion of "squat on a businessman", "reporter" is selected for the first
If the selected frequency is high, "reporter ride reporters" ineligible conversion results will be obtained. Therefore,
The use (combination) of the methods is desired. However, this method is often used in machine translation, but when a processor having relatively low processing capability is used, a problem arises in that the processing load increases and the speed required for conversion decreases.
As a result, a current document creation device often obtains a desired conversion candidate using the above-described method.

Words that are likely to appear in pairs are called "co-occurring words". The word pair (a pair of information of a preceding word and a word of a following word) is a word dictionary memory (which is referred to at the time of conversion, and includes additional information such as word reading, notation, frequency information, grammar information, and the like). (A stored memory) or a table provided separately from the word dictionary. At the time of conversion, a co-occurrence relation for a certain word is referred to, and if a paired word exists in a conversion candidate of a following word (which may be separated), that word is preferentially output. For example,
When a co-occurrence relationship of "people-crying" (() indicates that it can be changed by utilization) is registered, " people are not violent" Cry "
Work is co-occurrence relation to the portion of the underline as, "people ringing violently
Ku output of incorrect conversion result as "is prevented. In this case, the words are grouped (human / animal, etc.), and information on which group is included in each word is added, and the co-occurrence relation is regarded as a relation between the word groups, and when the words coincide with each other, priority is given. There is also a technique.

However, there are cases where this alone gives priority to mistakes. For example, when the co-occurrence relationship of “people-cry” works on the conversion of “not a person → not a person”, an erroneous conversion result of “crying by a person” is output. Therefore, when registering a co-occurrence relation, a connection relation between two co-occurring words, for example, "particle"
(In the above example, "people-cry (ku)-
However, there is a method that utilizes co-occurrence relations only when the particles also match.

(Problems to be Solved by the Invention) However, when two words registered in the co-occurrence table and a connection relation (for example, a particle) between the two words match in the conversion candidates, the co-occurrence relation is activated and priority output is performed. In this case, the co-occurrence relationship can be used for only one example. For example, "people - crying (Ku) - is" If you have been registered, in the example of "are not vigorously until the person" does not work is the co-occurrence relation by the presence of the particle "to". Therefore, in order to ensure that the co-occurrence relationship works, for example, “people cry”, “people cry”, “people and cry”, “people” Until-crying, "people-crying," ... all must be registered, and the number of co-occurrence combinations that must be registered is extremely large.

In addition, when two or more co-occurrence relationships are registered between words having the same reading, for example, when there is a "reporter-stop (ru)" or "train-stop-", there is a problem in which one has priority. Also occurs.

The present invention solves the above problem by converting reading information into a desired character string by referring to a co-occurrence relationship between words (or between a word concept and a human / animal / machine). It is another object of the present invention to provide a document creation apparatus capable of applying co-occurrence relations more effectively and outputting a desired candidate from homophones with priority.

[Structure of the Invention] (Means for Solving the Problems) In order to achieve the above object, a document creation device according to the present invention has an input means for inputting reading information and a searchable word character string corresponding to reading. The configuration of the registered dictionary storage means and the conversion processing means for searching the dictionary storage means based on the reading information input from the input means to obtain a desired word string, includes a pair of a certain word and another word. Conversion processing means, comprising: storage means for storing information indicating the following, a predetermined connection pattern of both words when these words appear as a pair in a character string, and a connection pattern replaceable with the predetermined connection pattern. However, if a plurality of homonyms exist as a result of the search of the dictionary information by the dictionary storage means, the search output may be replaced with the predetermined connection pattern or the predetermined connection pattern. To check whether a word pair registered in the storage means having a proper connection pattern exists, and if so, output the word pair in preference to other homonyms of both words. It is characterized by.

Further, information indicating a pair of a certain word and another word, a default connection pattern of both words when both words appear as a pair in a character string, a connection pattern replaceable with the default connection pattern, and Storing in the storage means priority information for a plurality of patterns in which the word pair can appear;
The conversion processing unit, when a plurality of homonyms exist as a result of the search performed by the dictionary storage unit with respect to the reading information, can replace the predetermined connection pattern or the predetermined connection pattern in a search output. A check is performed to determine whether a word pair registered in the storage unit having a connection pattern exists.If a plurality of word pairs exist as a result of the check, a word pair having a higher priority is determined according to the priority information. There is provided a document creating apparatus characterized in that a selected word pair is output with priority over other homonyms of these words.

(Operation) With the above configuration, when a plurality of homonyms are output as a conversion result of the reading information, the document creation device of the present invention selects a probable word from among them and gives precedence to other homonyms. Can be output.

Example First, a basic concept of the present example will be described. When co-occurrence is used, attention is paid to a connection relationship between two words (a word may be classified by concept and a co-occurrence relationship between concepts may be used). In general, the connection relation between words in a co-occurrence relation is “,,,,,,,,,,,,,,,,,,,,,,,”,
,,,,,,, Etc., a co-occurrence relationship without a particle between the words of the nomenclature called "concatenation", and a particle from the word of the previous verb called "modification" There is no co-occurrence. If the above connection relations are used as they are as co-occurrence information, the co-occurrence relations that must be registered will increase, so particles that perform the same function will be grouped and registered as co-occurrence information in this group unit. To be able to work. For example, "bird" + "ha" +
If “sounds” is registered in the co-occurrence table and “totori is absent” is input, “bird sounds” is prioritized as a conversion candidate with reference to the co-occurrence relationship. But,
The co-occurrence relationship, "bird cries," "bird squeal (voice)", "bird
It is also rather "apply particle" is the same co-occurrence relationship in particle other than "as shown in the example of" rather than to the bird. " Therefore, "bird"
-The connection information in the co-occurrence of “sound”
However, group (classification) information such as, but not limited to, is registered. As for the particle "wo", "noun +
An example of co-occurrence of "intransitive verb" and "noun + wo + transitive verb" can be considered. Then, it is possible to apply an equivalent co-occurrence relationship even if "" is changed to "no, ga, ga, wo, till" particles. Even if it is changed to each particle of,,,,,, and, an equivalent co-occurrence relationship can be applied. Table 1 shows an example of particle classification in this embodiment. The classification is based on the classification based on the case of the particle (nominative case, object case, union case, etc.). However, the classification of each particle is not necessarily grammatically unique. Adjustments are made to obtain optimal conversion results.

When a word having a co-occurrence relationship is registered, information indicating the classification in Table 1 is added as connection information therebetween. And
In the document creating apparatus of the present embodiment, when reading information is converted by applying a co-occurrence relation, words that exactly match the preceding and succeeding words and their connection relations (particles) match the corresponding word according to the conventional co-occurrence relation. Is output with priority, but a certain degree of priority is given to a pattern that matches a permitted connection relation (particle) to output the word. Table 2 shows an example of the priority in this embodiment.

In the present embodiment, a case where a plurality of co-occurrence relationships act on the same reading information (conversion example) occurs. For this reason, the priorities in Table 2 are applied in applying the co-occurrence relation. At the time of checking the co-occurrence relation, the priority is calculated, the co-occurrence relation having the higher priority score is selected, and the corresponding word is selected from the homophone and output as a candidate. In the example of Table 2, the priority is set higher as the pattern of the particle (connection form) completely matches and the phrase distance at which co-occurrence acts is shorter. Although only the adjacent / non-adjacent is seen in the above example, the score may be subdivided according to the number of clauses sandwiched between the clauses to which the co-occurrence relation is applied.

FIG. 1 is a functional block diagram of a document creating apparatus to which the present embodiment is applied.

A processing unit (actually realized by a CPU, interface hardware, and a program for controlling these units) includes an input control unit 11, which performs data (key code data, command data) input processing from the keyboard 110;
A kana-kanji conversion unit 13 that converts reading information (a kana character string in this embodiment) received by the input control unit into a sentence mixed with kanji and katakana, and proofreads and edits a document created as a conversion result (change of homophone / Editing work unit 15, which performs selection / deletion / movement / copy of character strings, created documents,
Alternatively, the display control unit 17 that controls the display of the conversion result (candidate)
Is provided.

A reading input buffer 112 is provided under the control of the input control unit 11 and the kana-kanji conversion unit 13. The reading input buffer 112 stores the reading information kana character string received by the input control unit 11,
The stored reading information character string is analyzed by the Kana-Kanji conversion unit 13 and converted into a sentence mixed with Kanji / Katakana.

Under the control of the screen display unit 17, there is a display memory 172 that stores pattern data displayed on a display device (display) 170. The display memory 172 is connected to the CPU, and dot patterns corresponding to character codes are written.

The kana-kanji converter 13 converts the dictionary 3
0, a co-occurrence table 31, a particle classification pattern table 32, and a priority rule table 33 are used. The dictionary 30 is an independent word dictionary that stores independent words (words mixed with kanji and katakana words) corresponding to the readings of the words as indices, and is used when cutting solid input readings into phrases as conversion units. Attached word dictionary, extracted words-between attached words /
It consists of a grammar dictionary for checking the correctness of the grammatical connection between the disconnected phrases. Each word stored in the independent word dictionary is provided with additional information such as grammatical information such as part of speech and inflection (used at the time of grammar check) and usage frequency information. The basic kana-kanji conversion method for obtaining a conversion candidate (homophonetic) from a dictionary configuration and reading information is described in, for example, Nikkei Electronics 1983.8.29, pages 180 to 215 and the section "Japanese processing". Further, since the method described in the reference can be applied, detailed description is omitted. In the co-occurrence table 31, information of two words having a co-occurrence relation (dictionary number of each word) and information indicating a connection relation of the two words (most appropriate connection relation-particle, concatenation, modification, etc.) are registered. . Note that the contents of the co-occurrence table may be stored in the dictionary 30 by storing a list of words that have a co-occurrence relationship with a certain word in the dictionary 30 and information of a connection relationship. Information registered in the co-occurrence table 31 is as follows.
Information of a category having a co-occurrence relationship may be used. For example,
“Cats, dogs, monkeys, etc.” are collectively classified into the category “animals” and registered as “animals-sounding-ga”. In this case, to each word stored in the dictionary 30 is added information on which category the word belongs to, for example, cat-animal, dog-animal, monkey-animal, i-human, her-human. You. The particle classification pattern table 32 is a table in which the connection relations that can be replaced with the connection relations registered in the co-occurrence table, that is, the co-occurrence patterns shown in Table 1 and a list of allowable connections are registered. Priority rule table 33
Is a table in which information is given as to how much priority is given between two words that are considered to have a co-occurrence relationship, that is, information on the point assignment shown in Table 2. The dictionary 30, the co-occurrence table 31, the particle classification pattern table 32, and the priority rule table 33
Is stored in the ROM connected to. If there is a possibility that additional registration will be performed later, the information may be stored in a readable / writable external storage device (magnetic disk).

The kana-kanji conversion unit 13 uses a co-occurrence pattern development table 130 and a homophone buffer 132 as a storage unit of the conversion result. The co-occurrence pattern development table is stored in the Kana-Kanji conversion unit 13
Among the conversion candidates, those which can apply the co-occurrence relationship among the conversion candidates (a plurality of homonyms) using the contents of the co-occurrence table 31, the particle classification pattern table 32, and the priority rule table 33 are described. When selected, the co-occurrence word pair and the priority score are stored in the co-occurrence pattern development table 130. The homophone buffer 132 stores homonyms output by the kana-kanji conversion unit 13 as a conversion result for the reading information in the reading information buffer 112, and readings of these homonyms. The conversion result includes a word in which a uniquely determined word (having no homonym) and a homonym are mixed. For the uniquely determined word, the kana-kanji conversion unit 13 writes the character code of the word read from the dictionary 30 to the document buffer 180. For a word having a homonym, a homonym buffer 132 storing a code indicating indefinite and a homonym of this word is stored.
Is written to the document buffer 180.

The editing work unit 15 proofreads and modifies the contents of the document buffer 180 in accordance with command input from the keyboard 110 (code input by pressing function keys such as move, copy, and character modification).
This is a program for editing. The editing work unit 15
In order to display the portion of the document to be edited, a character code string is read from the document buffer 180 and passed to the screen display unit 17. When the code indicating unconfirmed is read from the document buffer 180, the editing work unit 15 accesses the homophone buffer 132 based on the address information attached to this code, and And passes the code of the character string to the screen display unit 17. In addition, the homophone buffer 132
Each entry has a bit indicating a word to be displayed. This bit is turned on for a word to be displayed as a candidate in a homonymous word group of a certain word. The editing work unit 15 checks this bit and selects a word to be displayed as a candidate. Further, when the editing work unit 15 passes the character code of the candidate word to the screen display unit 17, the information that the word is an undetermined word is also added, and the screen display unit 17 confirms the display of the word. Indicate a different display (for example, overline, reverse, blink, etc.) Furthermore, the editing work unit 15
When there is an instruction from the keyboard 110 to display the next candidate of the homonym from the keyboard 110 (pressing the [next candidate] key), the word indicated by the cursor (line position of the cursor display) Performs the processing of displaying the next candidate. This is because the editing work unit 15 turns off the bit that is currently on in the homophone buffer 132, turns on the bit of the homonym in the next order, and repeats the display processing described above. Execute.

The screen display unit 17 displays the character codes sent from the editing work unit 15 and the modification instructions attached to the character codes (underlined, overlined,
A character pattern is generated based on (inversion, enlargement, etc.) and stored in the display memory 172. The basic pattern of characters is stored in a ROM (not shown), and can be read from the ROM according to codes. The contents of the display memory 172 are displayed on the display 170.

Next, the operation of the kana-kanji conversion unit 13 when utilizing the co-occurrence relationship will be described with reference to the flowchart of FIG.

For example, it is assumed that a solid character string of a solid sentence "Kisha shy shyness" is input as reading information and conversion is instructed (pressing the [Conversion] key on the keyboard 110) (step 10). . The read character string is stored in the buffer 112 as described above.

Upon receiving the conversion instruction, the kana-kanji conversion unit 15 separates the read character string into segments according to the contents of the dictionary 30, and converts the read character string into a sentence mixed with kanji (step 20). Since this conversion is the same as that of the conventional one, detailed description is omitted. Here, " Kishano / Kishiga / Kishani / Kishashi
And "(/ are numbered to distinguish. Explanation on word indicating a breakpoint of clauses.) And the word-separated, for words ~ homonym read from independent word dictionary of the dictionary 30 And is stacked in the homophone buffer 132. ~
For each reading of the independent word "Kisha", the homonym of the noun "Go back, your company, reporter, train" is stacked as a temporary conversion result. For "Kishashita", "coming home" is a combination of the sa variable verb "coming home" and the auxiliary verb "ta" indicating the past.
And the noun "kisha" when analyzing "ha" as the noun "lower".
Are stacked in the homonym buffer 132.

Next, the kana-kanji conversion unit 13 searches the co-occurrence table 31 and checks whether a pair of co-occurrence-related words registered in the co-occurrence table 31 exists in the conversion result (step 30). ). In the co-occurrence table 31, a) your company (noun) “no” reporter (noun) b) your company (noun) “no” train (noun) c) your company (noun) “ni” return to home (in) (intransitive verb) d ) Suppose that the co-occurrence relationship of the reporter (noun) “ga” returned to the office (do) (intransitive verb) e) the train (noun) “de” returned to the office (do) (intransitive verb) was registered. Word pairs that have co-occurrence relations in homonyms include: 1. Your company and reporter, 2. Your company and reporter, 3. Your company and reporter, 4. Your company and train, 5. Your company and train, 6. Your company And train, 7. Your company and return, 8. Reporter and return, 9. Train and return, 10. Your company and reporter, 11. Your company and reporter, 12. Your company and train, 13. Your company and train, 14. Your company and return, 15. Reporter and return, 16. Train and return, 17. Your company and reporter, 18. Your company and train, 19. Your company and return ), 20. Reporter and returning to home, 21. Train and returning to home, and these are stored in the co-occurrence pattern development table 130. This procedure is executed by the kana-kanji conversion unit 180 as follows. First, it is checked whether any homonyms of the words in the first phrase are registered as words preceding the co-occurrence pair. If there is a registered word, the homonyms of the words in the following phrases (in this embodiment, up to three phrases later so as not to increase the number of candidates) are compared with the preceding words. Checks for paired words.

When each of these pairs (21 types) is stored in the co-occurrence pattern development table 130, it is checked whether or not the “connecting relationship” between the words forming the pair matches (step 40). Flag information of "1" is added to those having the same "connection relationship" (step 4).
1). For the pair whose “connecting relation” does not match, access the particle classification pattern table 32 (step 4
3) It is checked whether the connection is "acceptable" (step 45). If the connection is "acceptable connection", a flag "2" is added (step 47).
If not, a flag of "3" is added (step 49), and the pair is added to the pair. Then, the kana-kanji conversion unit 13
Each time, the priority rule table 33 is accessed according to the numerical value of the flag (step 50), and the number of priority points is assigned based on the specified priority rule (step 60). This result is stored in the co-occurrence pattern development table 130 (step 8
0). The above operation is performed, and if it exists, it is stored in the co-occurrence pattern development table 130. The above operation is repeated in the order of the word in the second phrase and the word in the third phrase (steps 90 and 100), and all combinations (21 in the above example) are stored in the co-occurrence pattern development table 13.
Store in 0. Assuming that the analysis is completed up to the end of the sentence, in the above example, as a result, 1. your company and reporter 1-70 / kanichi, nominative, adjacent, 2. your company and reporter 1-50 / kanichi, nominative, non-neighbour, 3 .Your reporter 1-50 / Kanichi, Nominal, non-neighbour, 4.Your company and train 1-70 / Kanichi, Nomi, non-neighbour, 5.Your company and train 1-50 / Kanichi, Nomi, non-neighbour, 6.Your company and train 1-50 / Kanichi, Nominal, non-neighbour, 7.Your company and return to home (Shi) 3-10 / None, word, non-neighbour, 8.Reporter and home (Shi) 3-10 / Mischief, word, non-neighbour, 9. Train and return home (S) 3-10 / non-one, word, non-neighbour, 10. Your company and reporter 3-20 / non-one, body, adjacent, 11. Reporter 3-10 / None, Nominal, Not Neighbor, 12.Your Company and Train 3-10 / None, Nomi, No Neighbor, 13.Your Company and Train 3-10 / None, Nominal, No Neighbor, 14.Your Company 2-40 / permitted, word, non-neighbour, 15. Reporter and returned 1-50 / kanichi, word, non-neighbour, 16. Train and home 3-10 / Mischief, decree, Unnecessary, 17. Your company and reporter 3-20 / None, Nominal, adjacent, 18. Your company and train 3-20 / None, Nominal, adjacent, 19. Your company and returning to the company 21. Co-occurrence with 21 as follows: 20. Reporter and return home (Sh) 3-20 / None, Judgment, Adjacent, 21. Train and return home (S) 3-20 / N / A, Negative, Adjacent The homonym pair of the relationship is registered in the development table 130 together with the priority score.

For each homonym of each word, accumulate the score of the pair to which it applies (it does not matter whether it is before or after). ・ About the word, "your company" → 350 , "reporter" → 10,
Scores of “train” → 10 ・ About words, “reporter” → 120 , “train” → 10 ・ About words, “your company” → 110 , “reporter” → 90,
“Train” → 80 ・ Single, “Return to home” → 240 , “Reporter” → 8
0, “train” → 80 is obtained, and the highest priority among homonyms of each word is selected as the first candidate. That is, when a homonym group is stacked in the homonym buffer 132, the bit of the selected candidate is turned on so that the first candidate is displayed (step 110. In this case, the first candidate is stacked first). You can rearrange the order of the homonyms in the order of priority, which translates to "Your reporter has returned to your company." This result is grammatically and semantically sound. This indicates that it is not necessary for the operator to operate an unnecessary [next candidate] key to obtain a desired conversion result.

 Further, the present embodiment will be verified with another example.

A reading character string of "sensei is happy" is input, and "sensei (ga) / happy (shi)"
(Ta) ". Also, co-occurrence table
It is assumed that only the co-occurrence relationship between the teacher “ha” announcement (do) and the intransitive verb is registered in 31.

The "connecting relation" that teacher + announced + does not completely match the particle "ha" in the co-occurrence table, but the admissible connecting relation includes "ga" in the group, so the adjoining particles It is determined that it is permitted, and the priority score is a word, so 60 points are given. This time, there is no other combination that assigns a priority score using the co-occurrence table, so that no other homonym has a priority score at all.

Therefore, the predecessor + has announced + The tyranny + has announced + The oath + has given priority over the + announcement, and “Teacher + has announced +” is output. This is also a grammatically and semantically sound conversion result.

In the embodiment, in the dictionary 30, if each word registered corresponding to the reading is added and stored with a word paired with the word and its connection information, the co-occurrence table 31 is stored. It can be omitted.

When a candidate having the same priority score appears, the order may be determined in consideration of another measure such as the frequency of use. Further, it is also possible to exclude a candidate having a certain score or less. In this case, the operator may be able to set a score threshold.

Further, when utilizing the co-occurrence relationship between the categories to which the words belong, the classification code of each word output as the homonymous word in the step 20 may be obtained from the dictionary 30.

In addition, the priority score is further subdivided, and the degree of separation of phrases, the suitability of the nomenclature to the deep case (nominative case, possessive case, purpose case, etc.) for the term, and the degree of similarity (allowance) of the "connection relationship" May be added to change the scoring.

[Effects of the Invention] According to the present invention, since co-occurrence relations are applied to words between registered co-occurrence relations in consideration of the similarity of connection relations between the words, the words are actually registered. Optimal conversion results can be obtained for more patterns than the number of co-occurrence relationships.

We also prioritize the application of co-occurrence relationships,
When a plurality of co-occurrence relations simultaneously affect the conversion result, it is possible to output the optimum one as the conversion result.

Further, since the co-occurrence relation can be applied not only between adjacent words but also between distant words, the co-occurrence relation can be appropriately applied even if a modifier or the like exists between the target words.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a document creating apparatus to which the present embodiment is applied, and FIG. 2 is a flowchart showing an operation of a kana-kanji conversion unit in the embodiment. 13 Kana-Kanji conversion part 31 Co-occurrence table 32 Particle classification pattern table 33 Priority rule table 130 Co-occurrence pattern development table

Continued on the front page (72) Inventor Toshio Miyama 2-9-9 Suehirocho, Ome City, Tokyo Inside the Toshiba Ome Plant Co., Ltd. (72) Inventor Masanori Iwaki 2-9-9 Suehirocho Ome City, Tokyo Toshiba Ome Plant Co., Ltd. (56) References JP-A 1-229367 (JP, A) JP-A 59-2125 (JP, A) JP-A 60-124774 (JP, A) JP-A 61-75467 (JP, A) JP-A-63-98068 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G06F 17/21

Claims (2)

(57) [Claims] An input unit for inputting reading information; a dictionary storage unit in which a word character string is registered in a searchable manner in correspondence with the reading; and a dictionary storage unit based on the reading information input from the input unit. In a document creating apparatus having a conversion processing means for obtaining a desired word string by searching, information indicating a pair of a certain word and another word, a default of both words when both words appear as a pair in a character string And a storage unit for storing a connection pattern that can be replaced with the predetermined connection pattern. The conversion processing unit searches the dictionary storage unit for the reading information and finds that a plurality of homonyms are If there is, in the search output, there is a word pair registered in the storage means having the predetermined connection pattern or a connection pattern that can be replaced with the predetermined connection pattern. Whether to inspect, when present, document creation apparatus and outputting the word pair in preference to other homonym both words. 2. An input means for inputting reading information, a dictionary storage means in which a word character string is registered so as to be searchable corresponding to the reading, and the dictionary storage means based on the reading information input from the input means. In a document creating apparatus having a conversion processing means for obtaining a desired word string by searching, information indicating a pair of a certain word and another word, a default of both words when both words appear as a pair in a character string A connection pattern that can be replaced with the default connection pattern, and storage means for storing priority information for a plurality of patterns in which the word pair can appear, wherein the conversion processing means When a plurality of homonyms exist as a result of the search performed by the dictionary storage unit, the predetermined connection pattern or the predetermined connection pattern can be replaced in the search output. Whether there is a word pair registered in the storage unit having a proper connection pattern, and if a plurality of word pairs exist as a result of the inspection, a word pair having a higher priority is determined according to the priority information. And outputting this word pair with priority over other homonyms of these words.