JP2000187661A - Character string inputting device and its method and recording medium for recording character string input program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To quickly retrieve and input highly precise predicted candidates corresponding to the contents of a text at the time of retrieving proper predicted candidates whose input after the inputted character string is predicted from a predicted candidate dictionary at the time of inputting a character string. SOLUTION: This character string inputting device is provided with a predicted candidate dictionary 8 for storing predicted candidates constituted of semantic candidates obtained by classifying a character string for each semantic relation and example candidates obtained by classifying the character string for each cooccurence relation and priority for preliminarily setting which of the semantic candidate and the example candidate should be preferentially retrieved, a character string inputting part 2 for inputting a character string, a predicted candidate retrieving part 10 for retrieving the predicted candidates whose input after the character string inputted by the character inputting part is predicted from the predicted candidate dictionary based on the priority, a display part 5 for displaying the inputted character string and the retrieved predicted candidates, and a predicted candidate selecting part 3 for selecting one of the displayed predicted candidates as a character string to be inputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is applied to a word processor, a computer having a document processing function, an information processing apparatus, and the like, and is based on a character string that has already been input, and is predicted to be subsequently input. The present invention relates to a character string input device and a character string input method for predicting and displaying a character string, and a recording medium recording a character string input program.

[0002]

2. Description of the Related Art Conventionally, when a sentence is input to a word processor, an information processing apparatus, or the like, Japanese Patent Application Laid-Open No. 7-271774 describes a conventional technique for reducing input load by utilizing data already input by a user. Has proposed a sentence input device that predicts and displays the next appearing word (candidate) using the semantic information of the already input character string. This sentence input device is a method of comparing the similarity between a meaning vector of a sentence and a meaning vector of a word described in a meaning dictionary, predicting the next word, and displaying the predicted word in a selectable manner.

As another prior art, Japanese Patent Laid-Open No. 7-3
According to the description of Japanese Patent No. 34499, there is proposed a character string input device that predicts and displays a character string immediately following a character string that has been input halfway during text input.
This character string input device uses a method of searching a word dictionary for a word that matches a preamble, a method of using connection establishment of characters and characters, a method of using co-occurrence probability of words and words, and using a connection probability of characters. Based on the method and the part of speech information of the word, a method that includes a word that is easily connected immediately after that as a predicted character string is included.

[0004]

SUMMARY OF THE INVENTION Therefore, Japanese Patent Application Laid-Open No. 7-2
As described in Japanese Patent No. 71774, a method using a semantic vector, that is, general-purpose semantic information in which the degree of association between words is set in advance (for example, the attribute information and topic information of the word itself) is used. The method of determining has the advantage that it is possible to predict candidates according to the input text content and to follow changes in text content.

[0005] Further, since a plurality of related words can be grouped by one semantic information, there is an advantage that a necessary data capacity can be suppressed. For example,
If there are 100 words that are likely to appear at the same time in a sentence related to a certain topic (semantic relationship), one common topic label may be set as semantic information. The data volume is 10
0 × 1 = 100 words.

For example, when inputting a series of character strings, if individual input character strings are viewed locally, it is difficult to always obtain accurate prediction candidates by processing based on only one piece of semantic information. For example, when inputting a sentence such as "... Aerospace Agency's speed and progress of the planetary spacecraft ...", since the topic (semantic information) of the entire sentence is related to "space", the semantic information of the sentence is Based on the "planet, sun, gravity,
Rockets, satellites, ... ".

However, when attention is paid to the final word "progress" of the input character string, the candidates for directly connecting to the word "progress" include the connection such as the co-occurrence example candidate "process, situation, direction". Strongly, the content of this sentence is not based on semantic information. Therefore, immediately after the word “progress” is input, it can be said that a higher-precision candidate can be obtained by giving priority to an example candidate based on the co-occurrence relationship between words, rather than a semantic candidate based on semantic information.

On the other hand, as described in Japanese Patent Application Laid-Open No. Hei 7-334499, a method using word co-occurrence can predict fairly accurate candidates even for individual input words. On the other hand, a large amount of co-occurrence information (surface information) is required to cover all of the text input of any topic. For example, in the above input example, “Aerospace Station”, “Planetary Spacecraft”, “Spacecraft Speed”
It is necessary to comprehensively register the example candidates of each co-occurrence relationship, such as “speed and progress” and “travel and direction”.

For example, if there are 100 words that are likely to appear simultaneously in a sentence on a topic,
The data volume of all combinations of two co-occurring words is
₁₀₀ C ₂ = 100! ÷ (2! × 98!) = 4950 words. Therefore, there is a problem that the storage capacity of the candidate data of the co-occurrence information becomes large as compared with the candidate data based on the semantic information, and the cost of the memory is increased and the search processing speed is reduced. Therefore, a practical character string input device cannot be realized.

The present invention has been made in consideration of the above circumstances. For example, in inputting a character string, an appropriate prediction candidate predicted to be input after the input character string is obtained from a prediction candidate dictionary. When searching, by predicting the meaning candidates related to the meaning of the character string and predicting the example candidates in the co-occurrence relation between the character strings, the storage capacity of the prediction candidates stored in the prediction candidate dictionary can be increased. Provided is a character string input device, a character string input method, and a storage medium storing a character string input program capable of quickly searching for and inputting a more accurate prediction candidate corresponding to the content of a sentence without having to use the same.

[0011]

According to the present invention, there is provided a prediction candidate comprising a semantic candidate in which a character string is classified according to a semantic relationship and an example candidate in which a character string is classified according to a co-occurrence relation, and either of the semantic candidate or the example candidate. A prediction candidate dictionary that stores a priority set as to whether to preferentially search for a character string, a character string input unit that inputs a character string, and a prediction that a character string input by the character string input unit will be input. A prediction candidate search unit for searching the prediction candidate to be searched from the prediction candidate dictionary based on the priority, a display unit for displaying the input character string and the searched prediction candidate, and one of the displayed prediction candidates. And a prediction candidate selection unit that selects a character string as a character string to be input.

In the present invention, the prediction candidate dictionary may be constituted by a storage device such as a ROM, a floppy disk (FD), a hard disk (HD), and a mini disk (MD). The character string input unit and the prediction candidate selection unit may be configured by input devices such as a keyboard, a mouse, and a pen / tablet. The prediction candidate search unit may be configured by a computer CPU and a program code stored in a ROM or an HD.

According to the present invention, in inputting a character string,
When searching from the prediction candidate dictionary for an appropriate prediction candidate predicted to be input after the input character string, there is a co-occurrence relationship between the character strings while predicting a semantic candidate related to the meaning of the character string. By predicting an example candidate, it is possible to quickly search and input a more accurate prediction candidate corresponding to the content of a sentence without increasing the storage capacity of the prediction candidate stored in the prediction candidate dictionary. .

A kana-kanji conversion dictionary; and a character string conversion unit for converting a character string inputted with reference to the kana-kanji conversion dictionary into a kanji-kana mixed sentence, wherein the prediction candidate search unit includes A configuration may be such that a prediction candidate predicted to be input after the kanji kana mixed sentence converted by the unit is searched from the prediction candidate dictionary based on the priority.

In this configuration, the kana-kanji conversion dictionary is
For example, a storage device such as a ROM, a floppy disk, and a hard disk may be used. The character string conversion unit may be configured by, for example, a CPU of a computer and a program code stored in a ROM or an HD. According to this configuration, in inputting a sentence mixed with kanji kana, a more accurate prediction candidate corresponding to the content of the sentence can be quickly searched and input.

[0016] When the prediction candidate search unit matches a prediction candidate obtained by referring to the input character string and the co-occurrence relation, the prediction candidate search unit searches the example candidate more than the semantic candidate obtained by referring to the semantic relation. You may comprise so that it may be preferentially displayed on a display part.
According to this configuration, in inputting a character string, a more accurate prediction candidate corresponding to the content of a sentence can be searched and input at high speed.

The prediction candidate search unit may be configured to display a semantic candidate obtained by referring to the semantic relationship and a prediction candidate obtained by referring to the co-occurrence relationship on the display unit, respectively. Good. According to this configuration, since the semantic candidate group and the example candidate group are displayed so as to be identifiable, a more appropriate prediction candidate can be identified and input according to the content of the text.

The predictive candidate search unit, when there is a common predictive candidate between the semantic candidate obtained by referring to the semantic relationship and the example candidate obtained by referring to the co-occurrence relationship, searches the common predictive candidate. You may comprise so that it may display on a display part with priority over another prediction candidate. According to this configuration, a more appropriate prediction candidate can be obtained according to the content of a sentence.

A prediction candidate learning unit that stores the prediction candidates selected by the prediction candidate selection unit as learning candidates in a learning candidate storage unit; and a semantic candidate or co-occurrence obtained by the prediction candidate search unit by referring to a semantic relationship. When a learning candidate stored in the learning candidate storage unit exists in the example candidates obtained by referring to the relationship, a learning candidate search unit that displays the learning candidate on the display unit in preference to other prediction candidates is further provided. A configuration may be provided.

In this configuration, the learning candidate storage section includes, for example, a RAM, a floppy disk, a hard disk,
It may be constituted by a storage device such as an MD (mini disk). The prediction candidate learning unit and the learning candidate search unit may be configured by a CPU of a computer and a program code stored in a ROM or an HD. According to this configuration, the learning candidate (prediction candidate) stored in the learning candidate storage unit is displayed with a higher priority than other prediction candidates, so that a more appropriate prediction candidate can be obtained according to the content of the text. .

The semantic relationship referred to by the prediction candidate search unit is topic information stored in the prediction candidate dictionary, and the prediction candidate search unit determines a topic corresponding to the prediction candidate selected by the prediction candidate selection unit. Information may be configured to be displayed on the display unit. According to this configuration, when there is a meaning candidate predicted with reference to the topic information, the topic information to which the meaning candidate belongs is displayed on the display unit,
The flow of prediction processing based on topic information is clearly indicated to the user,
It becomes easier to select a prediction candidate.

The information processing apparatus further includes a topic information selection unit for selecting one of the topic information displayed on the display unit, wherein the prediction candidate search unit uses a character string input unit from a prediction candidate dictionary based on the selected topic information. It may be configured to search for a semantic candidate predicted to be input after the input character string.

In this configuration, the topic information selecting section may be constituted by an input device such as a keyboard, a mouse, a pen and a tablet. According to this configuration, the semantic candidate that is predicted with reference to the topic information and that matches the topic information selected by the topic information selecting unit is prioritized.
A highly accurate prediction candidate corresponding to the content of a sentence can be obtained.

According to another aspect of the present invention, a prediction candidate comprising a semantic candidate in which a character string is classified according to a semantic relationship and an example candidate in which a character string is classified according to a co-occurrence relationship, and any one of the semantic candidate and the example candidate Is stored in the prediction candidate dictionary, and a character string is input using the character string input unit, and is input after the character string input by the character string input unit. A predicted candidate is predicted from the predicted candidate dictionary based on the priority, an input character string and the retrieved predicted candidate are displayed on a display unit, and one of the displayed predicted candidates is displayed as a predicted candidate. A character string input method is provided, wherein the selection is performed using a selection unit.

[0025]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The embodiments of the present invention will be described below in detail with reference to the embodiments shown in the drawings. The present invention is not limited by this. The present invention will be described by taking a Japanese sentence input as an example of a character string input device that predicts and displays a character string that is predicted to be input subsequently based on a character string that has already been input.

FIG. 1 is a block diagram showing a basic configuration of a character string input device according to one embodiment of the present invention. In FIG. 1, reference numeral 1 denotes a central processing unit for executing each processing. The central processing unit 1 includes, for example, a CPU, a ROM, a RAM,
It is composed of a computer consisting of an O port. Reference numeral 2 denotes a character string input unit for inputting a character string and instructing conversion. The character string input unit 2 includes, for example, a keyboard, a mouse, a pen,
It is composed of input devices such as tablets and trackballs. Reference numeral 3 denotes a prediction candidate selection unit for selecting and instructing a prediction candidate. The prediction candidate selection unit 3 includes, for example, a character string input unit 2
And the same input device. 4 is a topic information selection unit for selecting and instructing topic information.
For example, it is configured with the same input device as the character string input unit.

Reference numeral 5 denotes a display unit for displaying the result of conversion of the input character string, prediction candidates, topic information, and the like.
Is composed of a display device such as a CRT display, an LCD (liquid crystal display), and a PD (plasma display). 6 is based on the control of the central processing unit 1,
The input / output control unit controls input / output signals of the character string input unit 2, the prediction candidate selection unit 3, the topic information selection unit 4, and the display unit 5. The input / output control unit 6 is, for example, a C
PU, ROM (Read Only Memory) and HD (Hard Dis
and k) the program code stored in the internal storage medium.

Reference numeral 7 denotes a kana-kanji conversion dictionary storing various dictionary information such as word reading, notation, and connection information of particles. The kana-kanji conversion dictionary 9 is stored in an internal storage medium such as a ROM or an HD. Data. 8
Is a prediction candidate dictionary that stores a co-occurrence relationship between words, an example candidate based on the word, a meaning relationship (topic information) to which the word belongs, a meaning candidate based on the word, and the like.
For example, it is composed of data stored in an internal storage medium such as a ROM or an HD.

Reference numeral 9 denotes a character string conversion unit for converting an input character string into a character string containing kanji and kana with reference to a kana-kanji conversion dictionary 7. The character string conversion unit 9 includes, for example, a CPU, a ROM, It consists of program codes stored in an internal storage medium consisting of an HD. Reference numeral 10 denotes a prediction candidate search unit that searches for a prediction candidate as a character string to be subsequently input based on a character string that has already been input, with reference to the prediction candidate dictionary 8. The prediction candidate search unit 10 includes, for example, It is composed of a program code stored in a computer CPU, an internal storage medium such as a ROM or an HD. Reference numeral 11 denotes a priority determination unit that determines the priority of a search criterion (semantic relationship or example relationship) when the prediction candidate search unit 7 searches for a prediction candidate.
U, a program code stored in an internal storage medium such as a ROM or an HD.

Reference numeral 12 denotes an input character string storage unit which functions as a buffer memory for storing a character string input from the character string input unit 2. The input character string storage unit 12 includes, for example, an RA.
It is composed of an internal storage medium composed of M (Random Access Memory) and FD (Floppy Disk). Reference numeral 13 denotes a converted character string storage unit which functions as a buffer memory for storing information relating to the conversion result of the input character string.
Reference numeral 3 denotes an internal storage medium including a RAM and an FD. Reference numeral 14 denotes a confirmed character string storage unit that functions as a buffer memory for storing information about the input and confirmed character strings. The confirmed character string storage unit 14 is configured by an internal storage medium including a RAM and an FD.

Reference numeral 15 denotes a prediction candidate storage functioning as a buffer memory for storing a character string relating to the prediction candidate.
The prediction candidate storage unit 15 is configured by an internal storage medium including a RAM and an FD. Reference numeral 16 denotes a learning candidate storage that functions as a buffer memory that stores the selected prediction candidate as a learning candidate.
It is composed of an internal storage medium composed of M and FD.

Reference numeral 17 denotes a prediction candidate learning unit that stores the selected prediction candidate in a learning candidate storage unit 16. The prediction candidate learning unit 17 includes, for example, a C
It is composed of a PU and program codes stored in an internal storage medium such as a ROM or an HD. Reference numeral 18 denotes a learning candidate search unit that searches for learning candidates from the learning candidate storage unit 16,
The learning candidate search unit 18 is, for example, a CPU of a computer.
And a program code stored in an internal storage medium such as a ROM or an HD. 19 is a control program,
It is a bus line for transferring various data such as addresses, and controls each unit via the central processing unit 1 and the bus line 19.

20 is an IC card composed of an EEPROM,
An external storage medium such as an FD, an MD, and a CD-ROM. The external storage medium 20 includes a character string conversion unit 9, a prediction candidate search unit 10, a priority determination unit 11, a prediction candidate learning unit 17,
A program code (character string input program) functioning as the learning candidate search unit 18 and dictionary data of the prediction candidate dictionary 8 may be stored.

By reading out the character string input program from the external storage medium 20 and installing it in the RAM of the central processing unit, the computer of the central processing unit 1 shares the semantic candidates and the character strings obtained by classifying the character strings for each semantic relationship. The prediction candidate dictionary 8 stores a prediction candidate composed of example candidates classified for each occurrence relationship and a priority set as to which of the meaning candidates or the example candidates is preferentially searched.
And inputting a character string from the prediction candidate dictionary 8 based on the priority, and searching for a prediction candidate predicted to be input after the character string input by the character string input unit 2. To realize a character string input device that can display a sequence and a searched prediction candidate on a display unit 5 and can select one of the displayed prediction candidates as a character string to be input using the prediction candidate selection unit 3. Can be.

FIG. 2 is a diagram showing the configuration of the internal storage medium according to this embodiment. In FIG. 2A, reference numeral 7a denotes kana-kanji conversion dictionary data including word readings and notations, 8a
Is a predictive candidate dictionary data composed of word co-occurrence relations, topic information to which the word belongs, 9a is a character string conversion program code for converting an input character string into a kanji kana mixed character string with reference to a word dictionary, and 10a is A prediction candidate search program code for searching a prediction candidate dictionary for a prediction candidate corresponding to a character string that has already been input, and 11a indicates a priority determination program code for determining the priority of the prediction candidate. 8a and control program codes 9a to 11a are stored in an internal recording medium such as a ROM or an HD.

In FIG. 2B, 12 is an input character string storage unit for storing an input character string, 13 is a converted character string storage unit for storing information relating to the conversion result of the input character string, Reference numeral 14 denotes a confirmed character string storage unit for storing information about input and confirmed character strings, and 15 denotes an area functioning as a prediction candidate storage unit for storing information about prediction candidates. Fifteen areas are configured on an internal recording medium including a RAM and an FD.

FIG. 3 is a diagram showing a configuration of a prediction candidate dictionary in which prediction candidates according to the present embodiment are stored based on semantic relationships. In FIG. 3, each word (prediction candidate) is configured with a value indicating the degree of association with each topic label. For example, the prediction candidate “objection” indicates that it is a semantic candidate related to the topic label W (topic information of the label W). The prediction candidate “satellite” is related to the topic information of the topic labels X and Y, and indicates that the degree of association with the topic information of the topic label Y is stronger than the topic information of the topic label X.

The topic information is a set of labels set in advance as indices representing the features of a sentence. Specifically, each topic label set in advance has a unique ID.
(Number). Then, in order to obtain the topic information of the input character string, prepare a word to which the value indicating the degree of relevance to each topic label is added, and collect the topic information of the word appearing in the confirmed character string. Then, the topic (semantic relationship) of the entire confirmed character string is determined.

FIG. 4 is a flowchart showing the procedure of the kana-kanji conversion process according to this embodiment. In FIG. 4, it is assumed that the display of the prediction candidates for the already input character string is performed in conjunction with the Japanese input such as the kana-kanji conversion process. S401: It is determined whether a character string for the conversion process has been input. In the case of YES, the process proceeds to step S402 and N
In the case of O, the process proceeds to step S407. S402: The character string is stored in the input character string storage unit 11.

S403: After storing the input character string, it is determined whether or not there is a conversion instruction for the conversion processing. In the case of YES, the process proceeds to step S404, and in the case of NO, the process proceeds to step S406. S404: The character string conversion unit 9 refers to the kana / kanji conversion dictionary 7 to perform conversion to a kanji / kana mixed character string, and stores the converted character string in the converted character string storage unit 13. This kana-kanji conversion processing may use a known method, and the structure of the kana-kanji conversion dictionary referred to is also known.
Detailed description is omitted. S405: Display the converted character string on the display unit 5.

S406: Waiting for an instruction to select a prediction candidate.
However, if there is no input character string in the initial state of step S401, the display of the prediction candidate is also in the initial state, and thus the state does not wait for the selection instruction of the prediction candidate. If yes,
After the prediction candidates are displayed on the display unit, if there is an instruction to select a prediction candidate, the process proceeds to step S407, and if NO, returns to step S401.

S407: Wait for an input such as a confirmation instruction or a next candidate request. It is determined whether a confirmation instruction has been issued. YES
In the case of, the selected prediction candidate is determined and processed as an input character string, and the process proceeds to step S408. In the case of NO, the process returns to step S401. In step S403, after the input character string is stored, even if there is no kanji conversion request, the system waits for a confirmation instruction.

S 408: The determined character string (for example, the notation code or the word ID of the determined character string) is stored in the determined character string storage unit 14. A topic label (topic information) related to each word is searched from the prediction candidate dictionary 10 using a word appearing in the confirmed character string as a keyword, and the confirmed character string storage unit 1 is searched.
4 is stored. S409: The prediction candidate search unit 10 updates the topic information of the input text based on the confirmed character string. S410: The prediction candidate search unit 10 checks whether a character string on a specific topic has been input. The related value (weight) is added to the topic label of the word in the determined character string, and a topic label exceeding a certain value is regarded as having a topic fired (hereinafter, such a state is referred to as a topic being fired).

S411: When a specific topic is ignited, the prediction candidate search unit 10 searches the prediction candidate dictionary 8 using the topic label as a keyword, and performs a prediction process based on the topic information. That is, a word related to a specific topic label is searched this time. S412: The prediction candidate search unit 7 determines whether there is a search candidate. In the case of YES, the process proceeds to step S413,
If NO, the process proceeds to step S414. S413: The prediction candidate search unit 10 stores the prediction candidate in the prediction candidate storage unit 15. In this case, the determination as to whether or not to be a prediction candidate is made when the degree of relevance of a topic is equal to or more than a certain value, and the value is determined according to how topic information is set.

FIG. 5 is a diagram showing the configuration of a prediction candidate dictionary in which prediction candidates according to the present embodiment are stored based on co-occurrence relationships. 501 is a preceding word (determined character string), 503 is a following word (example candidate) following the word 501, and 502 is a word 5
Particles that can co-occur between 01 and 503 are shown. For example,
If the last confirmed character string is "Air", searching this prediction candidate dictionary will return "Cargo, Control, Engineering, Transportation"
It is predicted that such words may co-occur. A dictionary that stores example candidates based on co-occurrence relationships between words is also used in kana-kanji conversion processing.

S414: The prediction candidate search unit 10 searches for an example candidate based on the co-occurrence relation of words in the prediction candidate dictionary, using the confirmed character string input immediately before as a keyword. S415: The prediction candidate search unit 10 determines whether or not an example candidate has been obtained as a result of the search for a prediction candidate having a co-occurrence relationship. In the case of YES, the process proceeds to step S416, and in the case of NO, the process proceeds to step S417. S416: The prediction candidate search unit 10 stores the searched example candidate in the prediction candidate storage unit 15.

S417: The priority determination unit 11 determines which one of the semantic candidate based on the semantic relationship (topic information, semantic information) or the example candidate based on the co-occurrence relationship (co-occurrence information, surface information) is prioritized. Determine the degree. S418: If there is a prediction candidate based on topic information and a prediction candidate based on co-occurrence information obtained simultaneously in a series of prediction processes, a prediction candidate based on co-occurrence information (example candidate)
Is displayed on the display unit 5 with priority. If there is a common prediction candidate based on the topic information and the co-occurrence information, the prediction candidate may be given the highest priority (for example, displayed first).

When there is no prediction candidate or only prediction candidates based on either topic information or co-occurrence information, a prediction candidate group based on semantic information and a prediction candidate group based on co-occurrence information are determined. Therefore, the priority is not determined. For the prediction candidate group based on the semantic information, the priority may be further determined from among them. For example, a method of giving priority to information having a high degree of relevance in topic information has been considered. Similarly, as for the prediction candidates based on the co-occurrence information, the words stored in the prediction candidate dictionary 8 having higher frequency values may be searched with priority.

When performing the kana-kanji conversion process,
When the learning candidate is in the prediction candidate in the conversion process, the search may be performed with priority. Note that the method of learning candidates and the method of searching for learning candidates are realized by known kana-kanji conversion processing, so detailed description is omitted.

FIG. 6 is a diagram showing the configuration of the learning candidate storage unit according to this embodiment. FIG. 6 shows an example in which a maximum of N prediction candidates to be learned are stored in the learning candidate storage unit 16. A portion 601 stores an ID number for uniquely identifying each word (learning candidate). For example, when the selected character string is a learning candidate in the input of the character string, the prediction candidate learning unit 17 receives the instruction of the learning candidate by the next candidate operation, adds the ID number, and adds the ID number. 16 is stored. The process of storing the learning candidates by the prediction candidate learning unit 17 may be performed immediately before or immediately after step S408 in FIG. In addition, the learning candidate storage unit 16
Can store the result of conversion processing such as kana-kanji conversion and the result of selection of prediction candidates in a mixed manner.

In the process of determining the priority of a prediction candidate, if there is a learning candidate, the learning candidate search unit 18 searches the learning candidate storage unit 16 using each prediction candidate as a keyword. Then, when there is a prediction candidate that matches the learning information, the learning candidate is displayed at the top, and is given priority over other prediction candidates. The learning candidate search processing by the learning candidate search unit 18 may be performed in the processing of step S417 in FIG.

FIG. 7 shows a display example 1 of prediction candidates according to the present embodiment.
FIG. In FIG. 7, reference numeral 701 denotes a part for inputting a character string, and 702 denotes a display part (display frame) of a prediction candidate. When the character string “aviation” is determined, as shown in FIG. 3, a prediction candidate having a co-occurrence relationship is searched using the determined character string “aviation” as a keyword. At this time, the topic information of the confirmed character string “aviation” is stored as the topic label X and its relevance 5 and the topic label Y and its relevance 4.

Regarding the firing condition of the topic information based on the confirmed character string, for example, in the confirmed character string, the firing condition that “if two or more words have a common topic label with a relevance of 4 or more, it is considered to be fired” is set. Suppose you set it. In this embodiment, the topic labels X and Y do not fire because there is no common topic label having a relevance of 4 or more in two or more words.
Therefore, a prediction candidate based on topic information cannot be obtained.

Next, as shown in FIG. 5, when a prediction candidate having a co-occurrence relation is searched using “aviation” as a keyword,
For example, each example candidate of “freight, control, engineering, transportation” is obtained. Reference numeral 702 in FIG. 7 shows a display example of these example candidates.

FIG. 8 shows a display example 2 of the prediction candidates according to the present embodiment.
FIG. As shown in FIG. 8, the input character string “space station is,” is determined, and based on the determined character string “space station is,” semantic candidates and example candidates are searched. Note that clause components such as the suffix “bureau” and case particle “ha” are excluded from the search target. Searching for semantic candidates based on topic relations using the fixed character string "space" as a keyword, Fig. 3
As shown in (1), the topic label Y is fired in association with the topic label Y, together with the topic information of “aviation” immediately before from the previous firing condition.

When a topic label that has fired is found in this way, topic information is searched for using the topic label Y as a keyword. "Space, satellite,
Observations, orbits, aviation, exploration, planets, etc. are available. Then, when there is no example candidate as a result of the search for the co-occurrence information, the prediction candidates are only the meaning candidates based on the topic information, and FIG. is there.

FIG. 9 is a display example 3 of the prediction candidates according to the present embodiment.
FIG. FIG. 9 shows a state in which the user has selected a required character string “planet” from the displayed prediction candidates, and, for example, the selected character string is identifiably displayed with a mark “◆”. FIG. 10 is a diagram illustrating a display example 4 of the prediction candidates according to the present embodiment. FIG. 10 shows a state in which the selected character string “planet” is captured as a fixed character string. In addition, the method of realizing the prediction candidate selection instruction and the determination instruction is not particularly limited.

FIG. 11 is a diagram showing a display example 5 of prediction candidates according to the present embodiment. In FIG. 11, the input character string “speed”
Shows the state immediately after the is determined. Example candidates are searched based on the fixed character string "speed". However, no example candidate is obtained because the part of the particle "ya" does not match, and only the semantic candidate "space, satellite, ..." based on topic information continues. It is displayed.

FIG. 12 is a diagram showing a display example 6 of the prediction candidates according to the present embodiment. FIG. 12 shows a state immediately after the input character string “progress” is determined. Searching for an example candidate based on the confirmed character string “progress” gives “situation, direction, obstruction”, and is displayed with priority over semantic candidates based on topic information. In addition, as shown in FIG. 12, an example candidate such as “aim” may develop and display the inflectional ending information. The expansion of the inflected endings and the connection after attachment can be easily realized by using various conversion dictionaries and tables used in the kana-kanji conversion processing.

FIG. 13 is a diagram showing a display example 7 of prediction candidates according to the present embodiment. FIG. 13 shows a state immediately after the input character string “progress” is determined, as in FIG. 12, but is an example in which semantic candidates based on topic information are displayed following example candidates based on co-occurrence information.

FIG. 14 is a diagram showing a display example 8 of the prediction candidates according to the present embodiment. In FIG. 14, reference numeral 1401 denotes a portion that displays example candidates based on co-occurrence information, and 1402 denotes a portion that displays meaning candidates based on topic information. As described above, the display of the prediction candidate is clearly distinguished from the example candidate and the semantic candidate, so that the display portion of the prediction candidate is prevented from being complicatedly changed. ), The input operation of the prediction candidate becomes easy. As described above, in the display examples 1 to 8 of the prediction candidates corresponding to the character strings for which the input character strings have been determined, the display regions of the prediction candidates are provided on the right side of the screen. However, the display method may be other display modes. .

FIG. 15 is a diagram showing a display example 9 of the prediction candidates according to the present embodiment. In FIG. 15, reference numeral 1501 denotes an indicator for displaying a topic label that is currently being fired,
It also serves as a topic information selection unit (selection instruction button) for the user to select an arbitrary topic label. Note that the highlighted topic label Y indicates that the topic label is currently igniting, and since the topic label Y is igniting, meaning candidates such as "space, satellite, orbit, ..." are displayed as prediction candidates. Have been. As a topic information selection instruction method, a new label may be fired when an unfired topic label is selected, and the firing may be canceled when a topic label that is being fired is selected.

FIG. 16 is a diagram showing a display example 10 of the prediction candidates according to the present embodiment. FIG. 16 shows a state in which the firing of the topic label Y has been canceled, a new firing instruction of the topic label X has been issued, and the candidate has been changed to a prediction candidate such as “sound speed, airport, aviation,...”. Note that the topic information may be selected at any timing at which an input operation is possible, such as after a character string is determined. In this way, if the topic information determined to be firing is incorrect, if you want to fire another topic information, if the text content being input changes, or if the input text is still When the topic does not ignite a little, the user can select an arbitrary topic label, and can search for and input a highly accurate prediction candidate corresponding to the content of the text.

[0064]

According to the present invention, when inputting a character string, when searching for a suitable prediction candidate that is predicted to be input after the input character string from the prediction candidate dictionary, the search is performed with respect to the meaning of the character string. By predicting example candidates that have a co-occurrence relationship between character strings while predicting semantic candidates to be performed, it is possible to respond to the contents of sentences without increasing the storage capacity of prediction candidates stored in the prediction candidate dictionary. Moreover, it is possible to quickly search for and input a prediction candidate with higher accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic configuration of a character string input device according to one embodiment of the present invention.

FIG. 2 is a diagram illustrating a configuration of an internal storage medium according to the present embodiment.

FIG. 3 is a diagram illustrating a configuration of a prediction candidate dictionary in which prediction candidates according to the present embodiment are stored based on a semantic relationship.

FIG. 4 is a flowchart illustrating a procedure of a kana-kanji conversion process according to the embodiment.

FIG. 5 is a diagram illustrating a configuration of a prediction candidate dictionary in which prediction candidates according to the present embodiment are stored based on co-occurrence relationships.

FIG. 6 is a diagram illustrating a configuration of a learning candidate storage unit according to the present embodiment.

FIG. 7 is a diagram showing a display example 1 of prediction candidates according to the embodiment.

FIG. 8 is a diagram showing a display example 2 of prediction candidates according to the embodiment.

FIG. 9 is a diagram showing a display example 3 of prediction candidates according to the embodiment.

FIG. 10 is a diagram showing a display example 4 of prediction candidates according to the embodiment.

FIG. 11 is a diagram showing a display example 5 of prediction candidates according to the embodiment.

FIG. 12 is a diagram showing a display example 6 of prediction candidates according to the embodiment.

FIG. 13 is a diagram showing a display example 7 of prediction candidates according to the embodiment.

FIG. 14 is a diagram illustrating a display example 8 of prediction candidates according to the present embodiment.

FIG. 15 is a diagram showing a display example 9 of prediction candidates according to the embodiment.

FIG. 16 is a diagram showing a display example 10 of prediction candidates according to the embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Central processing part 2 Character string input part 3 Prediction candidate selection part 4 Topic information selection part 5 Display part 6 Input / output control part 7 Kana-kanji conversion dictionary 8 Prediction candidate dictionary 9 Character string conversion part 10 Prediction candidate search part 11 Priority judgment Unit 12 input character string storage unit 13 converted character string storage unit 14 determined character string storage unit 15 prediction candidate storage unit 16 learning candidate storage unit 17 prediction candidate learning unit 18 learning candidate search unit 19 bus line 20 external storage medium

Claims (11)

[Claims] 1. A prediction candidate comprising a semantic candidate in which a character string is classified according to a semantic relationship and an example candidate in which a character string is classified according to a co-occurrence relationship, and which one of the semantic candidate and the example candidate is preferentially searched. A prediction candidate dictionary storing the set priority, a character string input unit for inputting a character string, and a prediction candidate predicted to be input after the character string input by the character string input unit. Candidate search unit that searches from the candidate dictionary based on the search, a display unit that displays the input character string and the searched prediction candidate, and a prediction that selects one of the displayed prediction candidates as the input character string A character string input device comprising: a candidate selection unit. 2. A kana-kanji conversion dictionary, and a character string conversion unit that converts a character string input with reference to the kana-kanji conversion dictionary into a kanji-kana mixed sentence, wherein the prediction candidate search unit includes: 2. The character string input according to claim 1, wherein a prediction candidate predicted to be inputted after the kanji kana mixed sentence converted by the character string conversion unit is searched from the prediction candidate dictionary based on the priority. apparatus. 3. The prediction candidate search unit, when the input character string matches a prediction candidate obtained by referring to a co-occurrence relationship, more than the meaning candidate obtained by referring to the semantic relationship. 2. The character string input device according to claim 1, wherein candidates are displayed with priority on the display unit. 4. The prediction candidate search unit displays a semantic candidate obtained by referring to a semantic relationship and a prediction candidate obtained by referring to a co-occurrence relationship on the display unit so as to be distinguishable from each other. The character string input device according to claim 1, wherein: 5. The prediction candidate search unit according to claim 1, wherein a common prediction candidate exists between the semantic candidate obtained by referring to the semantic relationship and the example candidate obtained by referring to the co-occurrence relationship. 2. The character string input device according to claim 1, wherein candidates are displayed on the display unit prior to other prediction candidates. 6. A prediction candidate learning unit that stores a prediction candidate selected by the prediction candidate selection unit as a learning candidate in a learning candidate storage unit, and a semantic candidate obtained by the prediction candidate search unit by referring to a semantic relationship. A learning candidate search unit for displaying, on the display unit, the learning candidate stored in the learning candidate storage unit in a case where the learning candidate stored in the learning candidate storage unit is present in the example candidates obtained by referring to the co-occurrence relation; The character string input device according to claim 1, further comprising: 7. A topic information stored in a prediction candidate dictionary, wherein the semantic relation referred to by the prediction candidate search unit is;
The character string input device according to claim 1, wherein the prediction candidate search unit displays topic information corresponding to the prediction candidate selected by the prediction candidate selection unit on the display unit. 8. A topic information selection unit for selecting one of topic information displayed on the display unit, wherein the prediction candidate search unit is configured to perform a search based on the topic information selected by the topic information selection unit. The character string input device according to claim 7, wherein a meaning candidate predicted to be input after the character string input by the character string input unit is searched from the prediction candidate dictionary. 9. A prediction candidate consisting of a semantic candidate in which a character string is classified for each semantic relationship and an example candidate in which a character string is classified for each co-occurrence relationship, and which of the semantic candidate and the example candidate are searched for preferentially. The set priority is stored in the prediction candidate dictionary, a character string is input using the character string input unit, and the prediction candidates predicted to be input after the character string input by the character string input unit are described above. Searching the prediction candidate dictionary based on the priority, displaying an input character string and the searched prediction candidate on a display unit, and selecting one of the displayed prediction candidates using a prediction candidate selection unit. Character string input method characterized by the following. 10. A computer preferentially searches for a prediction candidate composed of a semantic candidate in which a character string is classified for each semantic relationship and an example candidate in which a character string is classified for each co-occurrence relationship and which of the semantic candidate or the example candidate. Is stored in the prediction candidate dictionary, a character string is input using the character string input unit, and the prediction is predicted to be input after the character string input by the character string input unit. A candidate is searched from the prediction candidate dictionary based on the priority, the input character string and the searched prediction candidate are displayed on a display unit, and one of the displayed prediction candidates is input as a character string to be input. A storage medium storing a character string input program to be selected using a selection unit. 11. A prediction candidate consisting of a semantic candidate in which a character string is classified for each semantic relationship and an example candidate in which a character string is classified for each co-occurrence relationship, and which of the semantic candidate and the example candidate are searched for preferentially. A prediction candidate dictionary storing the set priority, a character string input unit for inputting a character string, and a prediction candidate predicted to be input after the character string input by the character string input unit. A prediction candidate search unit that searches from the prediction candidate dictionary based on the input candidate, a display unit that displays the input character string and the retrieved prediction candidate, and selects one of the displayed prediction candidates as the input character string. A document processing apparatus including a character string input device including a prediction candidate selection unit.