JP6221275B2 - Character input program and character input device - Google Patents

Description

  The present invention converts a read character string input from an operation unit into a predetermined notation character string, and converts the converted character string to an application running on the computer. The present invention relates to a program for functioning as a character input system that is determined as an input character string, and a character input device to which the program is applied.

  Japanese input systems built into small information terminal devices such as mobile phones have a function to convert a reading character string into a written character string according to a conversion operation, and a reading character string is changed according to a key operation. Each time, a function that extracts a candidate character string that matches the reading character string at the current time (hereinafter referred to as a “predictive conversion function”) or a character string that is input following the confirmed character string. For example, a function for predicting and displaying a character string that is likely to be displayed (hereinafter referred to as “connection prediction function”).

  Regarding the connection prediction function, Patent Document 1 associates a connection source character string (“Prediction Source Character String” in Patent Document 1) with a connection destination character string (“Connection Prediction Word” in Patent Document 1). It is described that a prediction dictionary is searched for a word that has been confirmed immediately before, and a linked prediction word having the determined word as a prediction source character string is extracted as a candidate.

  Regarding predictive conversion, Patent Document 2 sets a conversion dictionary having a configuration in which a plurality of types of attributes representing attributes suitable for an input situation are associated with a predetermined number of dictionary data, and input of reading characters is started. The input status at the time when the candidate is determined or when the candidate is confirmed, and when the candidate is extracted from the conversion dictionary and displayed, the candidate associated with the determined input status is higher It is described that the priority is raised so as to be displayed on the screen.

JP 2009-146201 A JP 2009-276818 A

  In the conventional connection prediction process, a connection destination character string with respect to a character string determined immediately before is extracted by a search for a dictionary in which a combination of a connection source character string and a connection destination character string is registered. However, it cannot be said that there is a high probability that a candidate intended by the user is called. In particular, case particles such as “GA”, “NO”, “NO”, etc. are used in connection with various words and phrases, so the user can only register the original words and phrases that are common to various words. It is difficult to increase the possibility that the phrase will be called.

  In the invention described in Patent Literature 1, a dictionary having a configuration in which a combination of a noun and a case particle is connected as a source character string and a destination character string is associated is set. It seems that the above problem can be solved. However, in order to increase the accuracy with which the target word is actually called, it is necessary to register the connection destination character strings for the combinations of a large number of nouns and case particles. If so, the capacity of the dictionary file becomes enormous, which is not realistic.

  For example, FIG. 3 of Patent Document 1 associates the word “clothes” with the predicted words “color”, “price”, and “price”. For sweaters, shirts, etc.), it is necessary to associate similar connection prediction words. In addition, phrases such as “color”, “price”, and “price” are used not only in the field of “clothes” but also in various items, and thus are connected to a large number of nouns representing these items. It is necessary to associate predicted words.

  Also in predictive conversion and kana-kanji conversion, the display order of candidates is generally determined based on the priority of each word stored in the dictionary and the past input history, so that the word intended by the user is not called. It may take time to input characters.

The present invention has been made paying attention to the above problem, and a first problem is to increase the possibility that a word intended by a user is called as a candidate.
Furthermore, the present invention increases the data capacity of the dictionary unit when information indicating the relationship between the two is linked to a combination of words that may be used in a row for the connection prediction process. The second problem is to improve the accuracy of candidate extraction without any problem.

  In a character input system constituted by a program according to the present invention, a dictionary part in which a combination of a read character string and a notation character string for each of a plurality of words is registered, and a read character string that receives input of a read character string from an operation part Input means, candidate extraction means for searching the dictionary part by the input reading character string or the confirmed character string and extracting a predetermined number of written character strings as input candidates, and input candidates extracted by the candidate extraction means Candidate display means for displaying a character string on the display section, and confirmation means for confirming a candidate character string selected in accordance with a candidate selection operation in the operation section as an input character string to the running application.

  In the dictionary part set by this program, a plurality of nouns are classified into a plurality of categories based on their superordinate concepts, and categories classified into various classified nouns and various words and phrases related to each category. Alternatively, attribute information indicating a related category is associated and registered. The candidate extracting means includes attribute information storage means for temporarily storing the attribute information of the noun in response to the character string including the noun associated with the attribute information being set as the input character string. At the same time, in the search for the dictionary portion in a state where the attribute information is stored in this means, the word / phrase associated with the stored attribute information is preferentially extracted.

  As described above, in the present invention, various nouns are classified into a plurality of categories based on the superordinate concept, and are registered in the dictionary unit in association with attribute information indicating the category to which each noun belongs. Various words related to each category are also associated with attribute information indicating the related category and registered in the dictionary unit. When a character string including a noun associated with attribute information is set as an input character string, the attribute information is stored in the attribute information storage means, and in subsequent searches for the dictionary part, the string information is stored in the stored attribute information. The attached phrases are preferentially extracted.

  Therefore, if you sufficiently pick up the words and phrases related to each category and set the attribute information, any nouns classified in the same category can call similar related words based on the attribute information common to them. It becomes possible. In addition, when registering information indicating a connection relationship between words for connection prediction in the dictionary unit, it is not necessary to set information indicating the connection relationship for each combination of words, and the data capacity of the dictionary unit is reduced. There is no fear of enlarging.

  In addition, in the candidate extraction process, even if the attribute information is stored in the attribute information storage unit, even if the phrase is not directly related to the character string including the noun that is the source of the storage, Since the phrase that is associated with the same attribute information as the noun is extracted with priority, the possibility that the phrase intended by the user is called as a candidate is increased.

In the above character input system, the dictionary unit includes information indicating a connection relationship between words that may be input before and after, and the candidate extraction unit includes the character string determined immediately before by the determination unit. It is possible to provide a function of performing a connection prediction process of searching the dictionary part and extracting a notation character string of a phrase having a connection to the confirmed character string as the next input candidate. The candidate extraction unit of the embodiment having this function, when performing the connection prediction process in a state where the attribute information is stored in the attribute information storage unit, the phrase having a connection to the character string determined immediately before The notation character string associated with the attribute information stored in the history information storage means is preferentially extracted.
According to this embodiment, as a phrase directly or indirectly connected to a noun while a series of text including the noun is input, it is possible to call a phrase related to the category of the noun as an upper candidate. .

  In another embodiment, the candidate extraction unit includes a dictionary unit that divides the reading character string into phrases and divides the reading character string into phrases according to the input of the reading character string having a length corresponding to a plurality of phrases. And a continuous phrase conversion function for setting a character string obtained by connecting the extracted character strings extracted for each phrase as a conversion candidate. Further, the candidate extracting means for executing the continuous phrase conversion function, when a notation character string including a noun linked to attribute information is extracted as a candidate by searching for any phrase in the reading character string, The attribute information is stored in the attribute information storage means, and the phrase linked to the stored attribute information is preferentially extracted in the search for the clause after the clause.

  According to the above embodiment, in continuous phrase conversion for a character string including a noun associated with attribute information, the accuracy of conversion of a phrase located after the noun can be increased.

  In the candidate extraction means of the above embodiment, each time the reading character string input by the reading character string input means changes according to the operation in the operation unit, the notation character string of the phrase that matches the changed reading character string is extracted. First conversion means (corresponding to the above-described predictive conversion), and second conversion means for extracting words that match the reading character string input at that time in accordance with the conversion operation in the operation unit ( For Japanese input, it corresponds to Kana-Kanji conversion). The above-mentioned continuous phrase conversion function can be provided in both of these two types of conversion means, but at least the second conversion means may be provided with the continuous phrase conversion function.

  Further, in the candidate extraction means of the present invention including each embodiment, the notation character string of the phrase linked to the attribute information stored in the attribute information storage means is confirmed as the input character string by the confirmation means. Accordingly, a function for deleting the attribute information stored in the attribute information storage means can be provided. According to this function, after a character string including a noun is set as an input character string, the attribute information that has been saved is obtained when a word related to the noun is called as a candidate, selected by the user, and confirmed. Is deleted, it is possible to prevent a word related to the previous category from being extracted as a candidate when a sentence related to another category is input after input of a sentence related to a certain category is completed. .

  The program according to the present invention can include all data constituting the dictionary part, but is not limited to this, only information for updating the existing dictionary part and candidate extracting means to a state where the present invention is applied. May be included.

  A computer in which a program according to the present invention is incorporated constitutes a character input device together with a connected operation unit and display unit. This character input device can be widely applied to not only portable information terminal devices such as mobile phones including smartphones and tablet terminals, but also general information processing devices such as ATMs and facsimile devices that have a poor character input environment.

  According to the present invention, for each of a plurality of categories set based on the superordinate concept of various nouns, the attribute information indicating the category is associated with a word belonging to or related to the category and registered in the dictionary unit, and the noun is included. When a character string is confirmed or is temporarily confirmed in the kana-kanji conversion process, a phrase related to the category of the noun is preferentially extracted as a candidate in a subsequent search for the dictionary part. The possibility that a word intended by the user is called up at the top of the candidate display screen is greatly increased, and the efficiency of character input can be improved.

  In addition, for connection prediction processing, even when information indicating the connection relationship between the two is combined with a combination of words that may be input before and after, a plurality of words belonging to the same category Since common information can be associated, the accuracy of candidate extraction by the connection prediction process can be increased without increasing the data capacity of the dictionary unit.

It is a functional block diagram of a character input device to which the present invention is applied. It is a table which shows the main structures of the dictionary data of the dictionary for conversion. It is a table which shows the main structures of the dictionary data of a connection prediction dictionary. It is explanatory drawing which shows the example of a transition of a character input screen. It is explanatory drawing which shows the example of a transition of a character input screen. It is explanatory drawing which shows the example of a transition of a character input screen. It is explanatory drawing which shows the example of a transition of a character input screen. It is a flowchart which shows the flow of a character input process. It is a flowchart which shows the detailed procedure of the connection prediction process in the flowchart of FIG. It is a flowchart which shows the detailed procedure of the prediction conversion process in the flowchart of FIG. It is a flowchart which shows the detailed procedure of the kana / kanji conversion process in the flowchart of FIG.

FIG. 1 shows a configuration example of a character input device to which the present invention is applied.
A character input device 100 according to this embodiment inputs a kana / kanji character string into a file that is installed in a smartphone and managed by a running application, and mainly includes a touch panel display 2 and a character input system (IME) 1. The configuration is as follows.

  The touch panel display 2 is provided on the front surface of the housing 3 of the smartphone (see FIGS. 4 to 7). The character input system 1 is software incorporated in a control unit (not shown) in the housing 3, and includes a keyboard processing unit 11, a display processing unit 12, a reading character string assembly processing unit 13, a candidate extraction unit 14, and a confirmation processing unit. 15, a conversion dictionary 16, a connection prediction dictionary 17, an attribute memory 18, and the like.

  The keyboard processing unit 11 displays a software keyboard 20 (see FIGS. 4 to 7) including a plurality of character keys 21 on the touch panel display 2 and accepts an operation on the keyboard 20. When the character key 21 is operated, the operation content is transferred from the keyboard processing unit 11 to the reading character string assembly processing unit 13, and the reading character string assembly processing unit 13 assembles a reading character string corresponding to the operation.

The candidate extraction unit 14 includes a prediction conversion function based on a reading character string and a kana-kanji conversion function, and a connection prediction function that predicts the next input character string from the confirmed character string.
In the predictive conversion, every time the reading character string assembled by the reading character string assembling processing unit 13 is changed, the notation character string of the word that matches the reading character string after the change is extracted from the conversion dictionary 16. In the kana-kanji conversion, in response to the operation of the conversion key 22 (see FIGS. 4 to 7) in the software keyboard 20, the notation character string of the phrase that matches the reading character string at that time is converted into the conversion dictionary. 16 is extracted.

  When the selection operation for the displayed candidate or the operation of the confirmation key 23 of the keyboard 20 is performed, the confirmation processing unit 15 confirms the selected candidate or the target character string for the confirmation operation as the input character string. In response to the confirmation, the candidate extraction unit 14 searches the connection prediction dictionary 17 using the character string confirmed by the confirmation processing unit 15 and extracts a notation character string of a word / phrase in which a relationship connected to the confirmed character string is set. Make predictions.

  The display processing unit 12 brings up a character input screen on the touch panel display 2 in response to the activation of the system 1, and displays the read character string assembled by the read character string assembly processing unit 13 and the candidates extracted by the candidate extraction unit 14. To do. When the input character string is confirmed by the confirmation processing unit 14, the display processing unit 12 also displays the input character string in a confirmed state.

In the conversion dictionary 16, combinations of reading character strings and notation character strings for each of many words are registered for predictive conversion and kana-kanji conversion. In the connection prediction dictionary 17, a large number of combinations of two words that may be input in series are registered.
Further, in order to improve the extraction accuracy of candidates, in the conversion dictionary 16 of this embodiment, a plurality of nouns are classified into a plurality of categories based on the respective superordinate concepts, and various types other than the nouns related to each category. Are also classified into related categories. Some words may be assigned to a plurality of categories.

  Each of the conversion dictionary 16 and the connection prediction dictionary 17 can be composed of one or a plurality of dictionary files. It is also possible to make a dictionary having a configuration in which these dictionaries 16 and 17 are integrated into one dictionary.

FIG. 2 shows a configuration example of the dictionary data of the conversion dictionary 16 limited to noun dictionary data. In this embodiment, seven categories of [place] [vehicle] [food] [drink] [media 1] [media 2] [clothing] are set, and the concept nouns suitable for each category are respectively set. Attribute information (specifically, the category name) indicating that it belongs to the category is set.
[Media 1] provides dynamic information such as video and music, and [Media 2] provides static information such as printed matter.

The dictionary data of the conversion dictionary 16 shown in FIG. 2 includes a reading character string in hiragana, a notation character string in kana kanji, and attribute information. For parts of speech other than nouns, dictionary data having a configuration similar to that shown in FIG. 2 is set. Among the registered words / phrases, there are words / phrases belonging to multiple categories such as “Frankfurt” and “Net” in FIG. 2, but attribute information is set because they do not belong to any category. There are no phrases (mainly adjuncts such as particles).
Although not shown in FIG. 2, regardless of the presence or absence of attribute information, the dictionary data of the conversion dictionary 16 includes candidates based on the part of speech information, the priority set by the system designer, and the actual confirmation status. Usage frequency information updated by the extraction unit 14 is included.

  Note that in the conversion dictionary 16 of this embodiment, only final forms are registered for words such as verbs and adjectives. In predictive conversion and kana-kanji conversion, for the words used in the conversion dictionary 16, the final form is extracted in accordance with the beginning of the reading character string, and various usage forms are used from the final form as necessary. Is derived and added to the candidates (see FIG. 7 described later).

  FIG. 3 shows a configuration example of the dictionary data of the connection prediction dictionary 17. In FIG. 3, only the case particles are used as the connection source words, but dictionary data having the same configuration is registered for other words.

  In the connection prediction dictionary 17 in this example, for a combination of words that are highly likely to be input consecutively, the read character string and the written character string of the connected word are associated with the written character string of the connected word. Registered dictionary data is registered. Further, when the connected word / phrase belongs to one of the above-mentioned seven categories, attribute information indicating the belonging category is included in the dictionary data. That is, when a word whose attribute information is registered in the conversion dictionary 16 becomes a connected word, the same attribute information is also registered in the dictionary data including the connected word in the connection prediction dictionary 17. become. Further, when a word / phrase for which no attribute information is set in the conversion dictionary 16 becomes a connected word / phrase, no attribute information is set in the dictionary data of the connection prediction dictionary 17.

  Further, each dictionary data of the connection prediction dictionary 17 includes standard priority and usage frequency information. Hereinafter, the frequency obtained by combining standard priority and usage frequency information will be referred to as “priority”.

  As shown in FIG. 2, in this embodiment, nouns are classified into categories corresponding to the superordinate concept of the concept indicated by the phrase. Also, as shown in FIG. 3, words that may be input directly or indirectly connected to nouns in each category are classified into the related categories.

  In FIG. 2, the range where the dictionary data of the noun belonging to [clothing] among the seven categories is located is surrounded by a thick frame, and the “fuku / clothing” dictionary used in the following specific examples therein Data is shown as a halftone dot pattern. In FIG. 3, dictionary data in which words and phrases belonging to [clothing] are set as connection destinations is shown by a halftone dot pattern.

  4 to 7 show examples of transition of candidate display on the character input screen when a sentence including “clothes” in [clothing] is input. In any case, the character input screen is divided into upper, middle, and lower three levels, the character input field 24 is set in the upper part, the candidate display field 25 is set in the middle part, and the software keyboard 20 is set in the lower part. The software keyboard 20 includes a total of 16 keys including a character key 21, a conversion key 22, and a confirmation key 23.

  FIG. 4 shows the transition of the character input screen when the reading character string “Fukuo” is input and converted into “clothing”. FIG. 4A is a screen immediately after the input of the reading character string “Fukuo” is completed, and a plurality of candidates extracted by predictive conversion are displayed in the candidate display column 25. FIG. 4 (2) shows a screen after “clothes” is selected in the candidate list, and “clothes” selected in the candidate list is displayed in the character input field 24 and is displayed as a candidate. The candidate list in the display field 25 is updated to display candidates extracted by connection prediction for the confirmed character string.

  As in this example, in the connection prediction when a character string composed of a combination of a noun and a case particle is determined, the connection prediction dictionary 17 is searched with the last case particle. That is, in the candidate display column 25 of FIG. 4B, candidates for words connected to “O” in “clothes” are displayed. Of these, the words “wear” and “buy” up to the sixth place are displayed. As shown in FIG. 3, the same [clothing] attribute information as “clothes” in front of “O” is set in “get”, “raise”, “borrow”, and “lend” (see FIG. 3). .

  FIG. 5 shows a case where the reading character string “Many” is inputted after the reading character string “Fukuo” is converted into “Clothing” by the same flow as FIG. 4 (1) and (2). Shows the transition of the character input screen. Due to space limitations, FIG. 5 omits the figure corresponding to FIG. 4 (1) and shows the screen from the point in time when “clothes” is confirmed (FIG. 5 (1)).

  As described above, when “clothes” is confirmed, the candidate display column 25 displays a candidate list in which words and phrases having the attribute information of [clothing] are arranged at the top (see FIG. 5 (1)). However, in the example of FIG. 5, the displayed candidate is not selected but a reading character is input, and the display of the candidate display column 25 is updated by the predictive conversion accompanying the input.

FIG. 5B shows a display example at the time when a reading character string “many” is inputted. In the candidate display column 25, only words having a reading of “many” are displayed, and words relating to [clothing] are not shown.
FIG. 5 (3) shows a display example when “many” is selected from the above candidate list and the connection prediction process is performed. In this embodiment, since each word / phrase registered as a connection destination word / phrase for “O” is also registered as a connection destination of “many”, the prediction of connection according to the selection of “many” again causes [clothing] A candidate list in which words having attribute information are positioned higher is displayed.

  6 also shows the transition of the character input screen when the reading character string is input after the reading character string “Fukuo” is converted to “clothes”, as in FIG. 5. Also in this case, “clothes” is selected from the candidates extracted by the predictive conversion for the reading character string “Fukuo”, and the display of the candidate display column 25 is performed by the connection prediction performed in accordance with the selection. However, the word / phrase in which the attribute information of [clothing] is set is updated to the candidate list arranged at the top (FIGS. 6 (1) and (2)).

  In the example of FIG. 6, next, the reading character “ka” is inputted (FIG. 6 (3)). Predictive conversion is performed by this input, and the display in the candidate display field 25 is updated to a candidate list with words that match the reading character “ka” in front. In this list, the attribute information of [clothing] is set. “Buy”, “borrow”, “lend”, and character strings based on these utilization forms the top.

FIG. 7 shows the transition of the character input screen in the case where kana-kanji conversion is performed in accordance with the operation of the conversion key 22 after the reading character string “Fukuki ni” is input.
FIG. 7A is an example of a screen immediately before the conversion key 22 is operated. In the candidate display column 25, the hiragana character string “Fukukiri” and Katakana character string extracted by predictive conversion are displayed. “Fukuwakiru” is displayed.

  FIG. 7B is an example of the screen after the conversion key 22 is operated. In this embodiment, the reading character string is decomposed and converted into “fukuo” and “kiru” by continuous phrase conversion, and the first candidate character string for each phrase is combined, so that Is displayed in the character input field 24 in an unconfirmed state. In addition, the conversion cursor is set to “unclothed” of this unconfirmed character string, and various conversion candidates for the first phrase are displayed in the candidate display column 25.

  In the examples of FIGS. 4 to 7, the noun “clothes” is entered, but this “clothes” is replaced with other nouns belonging to [clothing] (such as “shirt” and “jacket”). Even when an input operation similar to that in each case is performed, candidate displays other than the conversion result for the reading character string including the noun are the same as those in each case.

  As shown in the above examples, in the character input system 1 of this embodiment, in response to the determination of the noun for which attribute information is set, the noun determined in the subsequent conversion process and connection prediction process. Preferentially extract words with the same attribute information as. For this priority extraction, in this embodiment, the attribute information of the confirmed noun is stored in the attribute memory 18 until a word / phrase having the same attribute information is determined.

According to the above priority extraction, the possibility that the user's target word is displayed at the top is greatly increased, so that the labor required for character input can be reduced and the efficiency of the input work can be improved.
Further, in the connection prediction dictionary 17, by setting attribute information indicating the category for words that function as a connection destination to a plurality of words / phrases in the same category, the connection prediction dictionary 17 collectively associates the words / phrases with each connection source. Therefore, the convenience of input can be improved without increasing the data capacity.

  Furthermore, as shown in the example of FIG. 5, in this example, not only words that are directly connected to a character string by a combination of a noun and a case particle, but also words that are connected through other words are confirmed nouns. If there is a candidate for which the same attribute information is set, that candidate is extracted with priority. Even in Kana-Kanji conversion, conversion to a character string with a structure in which other clauses are located between a clause containing a noun for which attribute information is set and a clause containing a phrase for which the same attribute information as the noun is set is accurate. Can be implemented well.

FIG. 8 shows a flow of a series of processes performed in the character input system 1.
When the character input screen is launched and input of reading characters is started, a reading character string corresponding to the input is assembled (step S1), and predictive conversion processing based on the reading character string (details are shown in FIG. 10). ) Is executed (step S2), and candidates extracted by the predictive conversion are displayed (step S3). If the input of reading characters continues, step S4 becomes “YES” and the process returns to step S1.
Thereafter, while the input of the reading character string continues, the processes of steps S1 to S4 are repeated.

  When the conversion key 22 is operated at a predetermined time, step S5 becomes “YES” and the process proceeds to step S6, and kana-kanji conversion processing (details are shown in FIG. 11) using the reading character string at that time. To be implemented. Candidates extracted by the kana-kanji conversion process are also displayed by the process of step S3.

  When a candidate extracted by predictive conversion or kana-kanji conversion is selected, or when the confirmation key 23 of the keyboard 20 is operated, step S7 becomes “YES” and the process proceeds to step S8, and the character to be selected is selected. The character string being displayed in the column or character input field 24 is confirmed.

After the character string is confirmed, in step S11, after the connection prediction process (details are shown in FIG. 9) for the confirmed character string is performed, the extracted candidates are displayed (step S3). In this embodiment, before the connection prediction process, the attribute information of the confirmed phrase is checked (step S9). Here, when the phrase having the attribute information being stored in the attribute memory 18 is confirmed (step S9 is “YES”), the attribute memory 18 is reset in step S10, and the process proceeds to the connection prediction process. If a phrase that does not have stored attribute information is confirmed, the process skips step S10 and proceeds to a connection prediction process.
Even when the candidate extracted by the connection prediction process is selected, the processes in and after step S8 are performed according to the “YES” determination in step S7.

  Hereinafter, the connection prediction process, the prediction conversion process, and the kana-kanji conversion process will be described in detail with reference to FIGS. 9, 10, and 11 in order.

  First, in the connection prediction process shown in FIG. 9, it is checked whether or not a noun is included in the word / phrase determined immediately before (step S101). If no noun is included (step S101 is “NO”), an attribute is determined. It is checked whether attribute information is stored in the memory 18 (step S102). If the determined word includes a noun (step S101: “YES”), it is checked whether attribute information is set for the noun (step S104).

  When the phrase including the noun for which attribute information is set is confirmed as in the examples of FIGS. 4 to 6, step S101 and step S104 are sequentially “YES” and the process proceeds to step S105, where the confirmed noun Is stored in the attribute memory 18. Since the saving here is overwriting saving, when the attribute information is already stored in the attribute memory 18, the information is deleted.

  When the storage of the attribute information is completed, the process proceeds to step S103, and the connection prediction dictionary 17 is searched by the confirmed phrase (if the particle is located at the end of the phrase, only the letter indicating the particle), and the confirmed phrase is The connected words (that is, the connected words / phrases in the dictionary data based on the fixed words / phrases) are extracted preferentially from those having the attribute information being stored.

  Even if the confirmed phrase does not include a noun, but attribute information is stored in the attribute memory 18 (step S101 is “NO” and step S102 is “YES”), the process proceeds to step S103. A phrase connected to the confirmed phrase is extracted with priority from those having attribute information being stored.

  In step S103, in addition to words having attribute information being saved, words having high priority among words / phrases for which attribute information is not set are extracted. Therefore, when a phrase having attribute information being stored is not found, only a phrase that is not related to the attribute information is extracted.

  If no words are included in the confirmed word and no attribute information is stored (both steps S101 and S102 are “NO”), the noun is included in the fixed word, but attribute information is set for the noun. If not (step S101 is “YES” and step S104 is “NO”), the process proceeds to step S106. In step S106, the words and phrases connected to the confirmed words and phrases are extracted in descending order of priority regardless of the presence / absence or type of attribute information.

  According to the above processing, as shown in the example of FIG. 4, immediately after the phrase including the noun is confirmed, the phrase having the same attribute information as the noun can be called as a candidate. After a phrase including a noun is determined, a phrase having attribute information different from that of the noun or a phrase having no attribute information is determined, and a phrase candidate having the same attribute information as the noun can be called.

Reference is now made to FIG.
Even in the predictive conversion process, first, it is checked whether information is stored in the attribute memory 18 (step S201). If the attribute information is stored (“YES” in step S201), the process proceeds to step S202, the conversion dictionary 16 is searched for the read character string, and a word that matches the read character string in the forward direction is being stored. Are extracted with priority from those having the attribute information.

  If the attribute information is not stored in the attribute memory 18 ("NO" in step S201), the process proceeds to step S203, and a word that matches the reading character string in front is read regardless of the presence or type of attribute information. Extracted in descending order of match and priority to character string.

  Even in step S202, if there are a predetermined number or less of words / phrases having attribute information being stored, words that have no attribute information set and that match the reading character string in front are sorted in descending order of priority. Extract.

  In both steps S202 and S203, when the final form of a usage word such as a verb or adjective is extracted by a forward matching search with the reading character string, the final form of the final form is changed while maintaining the range of the matching. Thus, a plurality of conversion candidate character strings are generated. The candidate list shown in FIG. 6 (3) is called in step S202 to which this function is applied.

Next, with reference to FIG. 11, a detailed procedure of kana-kanji conversion processing will be described on the assumption that continuous phrase conversion is performed.
First, in step S301, after the reading character string is cut into a plurality of phrases based on rules registered in advance, the process proceeds to a process of extracting candidates for the first phrase.

  In this candidate extraction process, it is checked whether or not the attribute information is stored in the attribute memory 18 (step S302). If the attribute information is stored (step S302 is “YES”), the conversion dictionary 16 starts with the head. Are extracted in preference to those having attribute information being stored. On the other hand, if the attribute information is not stored in the attribute memory 18 (step S311 is “YES”), words that match the first phrase are extracted in descending order of priority.

  Note that “words that match the first phrase” in steps S303 and S304 and later-described steps S310 and S311 usually mean words that match the reading character string, but the endings of verbs and adjectives change. As for the phrase to be used, the phrase whose reading character string in the stem portion matches the first phrase is also included in the “term that matches the first phrase”. For the latter phrase, candidates for matching the reading character string are generated by changing the endings while leaving the matching range for the reading character string of the first phrase.

  In the candidate extraction process for the first phrase, the first candidate among the extracted candidates is provisionally determined. In step 305, it is checked whether or not a noun having attribute information is included in the first candidate. If a noun having attribute information is included, the attribute information is stored in the attribute memory 18 in step S306.

  Hereinafter, candidate extraction processing for the currently focused phrase is performed by paying attention to each phrase below the second phrase (steps S307 to S313). Also in this case, the presence or absence of attribute information in the attribute memory 18 is checked (step S309), and if the attribute information is stored, the phrase that matches the phrase being focused on has the attribute information being stored. (Step S309 is “YES” and step S310 is executed), and if the attribute is not stored, the phrase that matches the phrase of interest regardless of the presence or type of attribute information Are extracted in descending order of priority (step S309 is “NO” and step S311 is executed).

  Also, in the process for the second phrase and below, if a character string including a noun for which attribute information is set is extracted as the first candidate (“YES” in step S312), the attribute information is stored in the attribute memory 18 in step S313. Save.

  Note that also in step S313 and the previous step S306, when other attribute information is already stored in the attribute memory 18, the previous attribute information is deleted by overwriting. In steps S303 and S310, a predetermined number of words / phrases for which attribute information is not set are extracted in addition to words / phrases having attribute information being stored.

  When the processing for all the phrases is completed, step S307 becomes “NO”, and the process proceeds to step S314. By concatenating the first candidates for each phrase, the converted character string is provisionally determined, and then the process is terminated. .

  According to the kana-kanji conversion process described above, even when a continuous phrase including a noun having attribute information is to be converted, a phrase having the same attribute information as the noun is extracted as the first candidate for the phrase after the noun. can do. Therefore, not only in the case where the noun having attribute information is located in the first phrase as in the example shown in FIG. 7, the character string (for example, “Cute / Clothing” in which the noun having attribute information is located in the second and subsequent phrases is used. ) /, And even if a word string with the same attribute information as the noun is separated from the noun (for example, “clothes / many / bought”), Can be batch converted.

  Although not shown in FIG. 11, after the above-described kana-kanji conversion is completed, an operation for correcting the conversion result for some phrases is performed, and the subsequent phrases may be converted again accordingly. Even in that case, the processing according to the procedure according to FIG. 11 can be performed.

  Also, in the predictive conversion process when a long reading character string is input, continuous phrase conversion can be performed by the same method as in FIG.

As described above, in the character input system 1 according to this embodiment, the attribute information is stored in the attribute memory 18 in response to the determination or provisional determination of the noun having the attribute information. A word or phrase having attribute information is preferentially extracted.
The stored attribute information is deleted in response to the user selecting and confirming the word / phrase in which the attribute information is set, and another noun having the attribute information is extracted and stored. Is deleted (step S105 in FIG. 9, steps S306 and S313 in FIG. 11). Therefore, when a topic moves to another category after a sentence indicating a topic related to a certain category is input, it is possible to prevent a concept word related to the previous topic from being extracted and It becomes possible to call as a candidate with higher accuracy.

  The character input system 1 is set by incorporating a program that constitutes the entire system into an information terminal device before shipment. In addition, the candidate extracting unit 14 and the conversion unit can be used in an information terminal device having an existing character input system. It can be set by incorporating update programs such as the dictionary 16 and the connection prediction dictionary 17.

  Moreover, this program can be introduced not only in an information terminal device having a touch panel display such as a smartphone or a tablet terminal but also in an information terminal device having a mechanical operation unit such as a conventional mobile phone.

DESCRIPTION OF SYMBOLS 1 Character input system 2 Touch panel display 11 Keyboard process part 12 Display process part 13 Reading character string assembly process part 14 Candidate extraction part 15 Confirmation process part 16 Dictionary for conversion 17 Connection prediction dictionary 18 Attribute memory 20 Software keyboard 24 Character input column 25 Candidate Display field 100 Character input device

Claims (6)

A computer to which an operation unit and a display unit are connected, a dictionary unit in which a combination of a read character string and a written character string for each of a plurality of phrases is registered, and a read character string input that receives input of the read character string from the operation unit Means, candidate extraction means for searching the dictionary portion with the input reading character string or the confirmed character string, and extracting a predetermined number of written character strings as input candidates, and the input candidates extracted by the candidate extraction means A candidate display means for displaying the character string on the display unit, and confirming the candidate character string selected according to the candidate selection operation on the operation unit as an input character string to the application running on the computer A program for functioning as a character input system comprising:
Multiple nouns are classified into multiple categories based on their superordinate concepts, and attribute information indicating categories or related categories is associated with each classified noun and various words related to each category. A program for setting the dictionary portion of the registered configuration in the computer is included,
The candidate extracting unit includes an attribute information storage unit that temporarily stores the attribute information of the noun in response to a character string including the noun associated with the attribute information being set as an input character string. in the search for a dictionary unit under a state where the attribute information is stored in this unit, extracting phrases that are linked to the attribute information that is the stored priority,
In the state where the candidate display means displays the words extracted by the candidate extraction means, when the reading character string input means accepts input of a reading character string, the candidate extraction means Among the displayed words and phrases, the words that match forward with the reading character string received by the reading character string input means are extracted, and if the extracted word or phrase includes a closing use word, the use A program for character input, characterized in that a phrase related to a utilization form other than the word end form is added to an input candidate . The dictionary portion includes information indicating a connection relationship between words that may be input in series before and after,
The candidate extraction unit searches for the dictionary part using the character string that was determined immediately before by the determination unit, and performs a connection prediction process that extracts a notation character string of a phrase that is connected to the determined character string as the next input candidate. In the case where the connection prediction process is performed in a state in which attribute information is stored in the attribute information storage unit, the function in the attribute information storage unit, the word in the phrase having a connection to the character string determined immediately before The program for character input according to claim 1, wherein the notation character string associated with the attribute information stored in the attribute information storage unit is preferentially extracted. The candidate extraction unit searches the dictionary unit by dividing the reading character string into phrases in response to the input of the reading character string having a length corresponding to a plurality of phrases. And a continuous phrase conversion function for setting, as a conversion candidate, a character string obtained by connecting the extracted character strings extracted for each phrase,
The candidate extracting means for performing the continuous phrase conversion, when a notation character string including a noun linked to the attribute information is extracted as a candidate by searching for any phrase in the reading character string, the attribute information is while stored in the attribute information storage unit, the extracted preferentially phrase tied to the stored attribute information in the search for clauses after the phrase, character input according to claim 1 or 2 Program. The candidate extracting unit extracts a notation character string of a word that matches the changed reading character string every time the reading character string input by the reading character string input unit changes according to an operation in the operation unit. 1 conversion means, and a second conversion means for extracting a phrase that matches the reading character string input at that time in accordance with the conversion operation in the operation unit, of these two types of conversion means The character input program according to claim 3 , wherein the continuous phrase conversion function is provided in at least the second conversion means. The candidate extracting means includes the attribute information storage means in response to a notation character string of a phrase linked to attribute information stored in the attribute information storage means being confirmed as an input character string by the confirmation means. Delete attribute information stored in
The program for character input described in any one of Claims 1-4 . A casing having an operation unit and a display unit, and a control unit incorporated in the casing, and a combination of reading character strings and notation character strings for each of a plurality of words / phrases are registered in the control unit. A dictionary part, a reading character string input means for accepting input of a reading character string from the operation part, and searching the dictionary part by the inputted reading character string or a confirmed character string and inputting a predetermined number of written character strings Candidate extraction means for extracting as candidates, candidate display means for displaying the character strings of the input candidates extracted by the candidate extraction means on the display section, and candidates selected according to the candidate selection operation in the operation section A character input device provided with a confirmation means for confirming a character string as an input character string,
In the dictionary unit, a plurality of nouns are classified into a plurality of categories based on the respective superordinate concepts, and each classified noun and various categories related to each category are respectively classified into categories or related categories. Attribute information is linked and registered,
The candidate extracting unit includes an attribute information storage unit that temporarily stores the attribute information of the noun in response to a character string including the noun associated with the attribute information being set as an input character string. in the search for a dictionary unit under a state where the attribute information is stored in this unit, extracting phrases that are linked to the attribute information that is the stored priority,
In the state where the candidate display means displays the words extracted by the candidate extraction means, when the reading character string input means accepts input of a reading character string, the candidate extraction means Among the displayed words and phrases, the words that match forward with the reading character string received by the reading character string input means are extracted, and if the extracted word or phrase includes a closing use word, the use A character input device characterized by adding a phrase related to a utilization form other than a word termination form to an input candidate .

Images