JP2011248538A - Symbol input support device, symbol input support method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically perform determination of a symbol in consideration of a content of sentence and insertion of the symbol in consideration of a role and a position following an actual usage, when inserting the symbol such as a pictograph into an input sentence.SOLUTION: A symbol input support device comprises: symbol appearance frequency learning means for calculating co-occurrence word information indicating frequency of appearance of each symbol in the sentences for learning with co-occurence words, context information, and co-occurence information of the symbol and the word in a predetermined dictionary given to each symbol, and storing a calculated result in learning result storage means as a learning result; and inserting symbol selecting means for calculating a sentence feature amount when inserting the symbol at each position of symbol inserting position candidates in input sentences for each insertion target symbol by using the learning result stored in the learning result storage means, specifying the symbol and the inserting position most appropriate to insert based on the sentence feature amount, and storing the symbol and the inserting position in conversion candidates symbol storage means.

Description

  The present invention relates to a symbol input support device, a symbol input support method, and a program for supporting insertion of symbols such as pictographs in an input sentence.

  2. Description of the Related Art Conventionally, a character input device such as a mobile phone or a personal computer that can easily input an illustration-like symbol character (pictogram) in a sentence.

  For example, Patent Document 1 automatically inserts character string candidates such as pictographs and punctuation that are likely to follow the end-of-sentence character string when a specific end-of-sentence character string such as “is” or “mas” is input. Discloses a character input device that saves the trouble of inputting sentence end characters.

  Further, Patent Document 2 has a storage unit that associates a pictograph with a predetermined character string, and converts the predetermined character string in the input character string into a specific pictogram registered in the storage unit, thereby converting the pictograph. A mobile terminal device that automatically inserts and saves the trouble of inputting pictograms is disclosed.

JP 2006-331066 JP 2010-15221 A

  However, in the technique described in Patent Document 1 specializing in input support for sentence end characters, there is a possibility that pictograms that do not take into account the content of sentences are displayed as input candidates. Further, the insertion position of the pictogram is also inserted only at the end of the sentence, and there is a possibility that the pictogram may not be used according to the actual usage of the pictogram that may be used in the sentence.

  In the technique described in Patent Document 2, since the pictogram is inserted after the word registered in the storage unit that appears in the sentence, it may not be in accordance with the actual usage of the pictogram appearing instead of the word or at the end of the sentence There is sex. In addition, since the insertion is performed by matching with a word, the contents of a part of the sentence are taken into consideration, but the contents of the whole sentence are not taken into consideration.

  The present invention has been made in view of the above problems, and when inserting a symbol such as a pictogram into an input sentence, the symbol is determined in consideration of the content of the sentence, and the role in accordance with the actual usage method It is an object of the present invention to provide a technique for automatically inserting symbols in consideration of positions.

  In order to solve the above problems, the present invention is a symbol input support device for inserting a symbol into an input sentence, and for each symbol in a sentence for learning, a symbol representing the frequency of appearance with a co-occurring word. Symbol appearance frequency for calculating word occurrence information, context information, and co-occurrence information of the symbol and a word in a predetermined dictionary given for each symbol, and storing the calculation result as a learning result in the learning result storage means For each symbol to be inserted for each position of the learning means and the symbol insertion position candidate in the input sentence, the learning result stored in the learning result storage means is used for each position of the symbol insertion position candidate. An inserted symbol that calculates a sentence feature when a symbol is inserted, identifies a symbol and an insertion position most suitable for insertion based on the sentence feature, and stores the symbol and the insertion position in the conversion candidate symbol storage unit Selection Configured as a symbol input support apparatus characterized by comprising: a stage, a.

The inserted symbol selection means includes co-occurrence word information, context information, and co-occurrence information with words in the predetermined dictionary when a symbol is inserted at the position of the symbol insertion position candidate in the input sentence. It is possible to obtain the symbol appearance probability based on the learning result as the sentence feature amount.
The symbol input support device may further include a character string conversion unit that inserts a symbol into the input sentence using the symbol specified by the insertion symbol selection unit and the insertion position.

  When the plurality of identical symbols are specified by the insertion symbol selecting means, the character string converting means is configured to use the plurality of identical symbols based on the sentence feature amounts corresponding to the plurality of identical symbols. The symbol can be selected and the symbol can be inserted at the insertion position corresponding to the symbol in the input sentence.

  The present invention may also be configured as a symbol input support method including steps corresponding to each process in the symbol input support device. Further, the present invention may be configured as a program for executing each step in the symbol input support method.

  According to the present invention, when a symbol such as a pictograph is inserted into an input sentence, the symbol is determined in consideration of the contents of the sentence, and the insertion of the symbol in consideration of the role and position in accordance with the actual usage is automatically performed. Can be performed.

It is a functional block diagram of the symbol input assistance apparatus 10 in embodiment of this invention. It is a figure which shows an example of a symbol list. It is a figure which shows an example of a text set. It is a figure which shows an example of a meaning dictionary. 4 is a flowchart for explaining an outline of the operation of the symbol input support device 10. 4 is a flowchart for explaining an operation of a symbol appearance frequency learning unit 3; It is a figure which shows the example of the co-occurrence word and the calculation result of co-occurrence relation. It is a figure which shows the example of the calculation result of co-occurrence relation with a right and left word. It is a figure which shows the example of the calculation result of the co-occurrence relationship for every part of speech of right and left. 5 is a flowchart for explaining the operation of an insertion symbol selection unit 4; It is a figure which shows the example of the symbol insertion position candidate when making between morphemes a symbol insertion position candidate. It is a figure which shows the example of a symbol insertion position candidate when a content word morpheme is made into a symbol insertion position candidate. It is a flowchart for demonstrating the symbol specific process for every morpheme division | segmentation which the insertion symbol selection part 4 performs. It is a flowchart for demonstrating the symbol specific process for every content word morpheme which the insertion symbol selection part 4 performs. 4 is a flowchart for explaining the operation of a character string conversion unit 5;

  Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings.

(Device configuration)
FIG. 1 shows a functional block diagram of a symbol input support device 10 for inserting a symbol in accordance with a context at an appropriate position of an inputted character string in the embodiment of the present invention.

  As shown in FIG. 1, the symbol input support apparatus 10 includes an input unit 1, an output unit 2, a symbol appearance frequency learning unit 3, an insertion symbol selection unit 4, a character string conversion unit 5, a learning data storage unit 6, and a learning result. A storage unit 7, a meaning dictionary storage unit 8, and a conversion candidate symbol storage unit 9 are provided.

  The input unit 1 is a functional unit for inputting a sentence (character string) to be inserted with a symbol. The output unit 2 is a functional unit for outputting a sentence in which a symbol is inserted. The symbol appearance frequency learning unit 3 uses the learning data stored in the learning data storage unit 6 to learn the co-occurrence relationship between symbols and other words, and stores the learning result in the learning result storage unit 7. It is a functional part to do. The insertion symbol selection unit 4 is a functional unit that specifies a symbol to be inserted into an input sentence and an insertion position using a learning result. The character string conversion unit 5 is a functional unit that inserts a symbol into the input sentence based on the symbol specified by the insertion symbol selection unit 4 and the insertion position, and outputs the sentence with the symbol inserted via the output unit 2. is there.

  The learning data storage unit 6 stores a text set and a symbol list as learning data. FIG. 2 shows an example of a symbol list. As shown in FIG. 2, symbols used in this embodiment are pictographs that can be input from a mobile phone or the like. Of course, it is possible to use symbols other than pictograms. The text set is a set of texts including symbols in the symbol list in one sentence. An example of a text set is shown in FIG.

  The learning result storage unit 7 stores the learning result by the symbol appearance frequency learning unit 3. The meaning dictionary storage unit 8 stores a meaning dictionary. The semantic dictionary is a dictionary used when calculating the relationship between a sentence including a symbol and the semantic meaning of the symbol, and includes a set of pairs of a symbol and a linguistic expression (word) representing the symbol. FIG. 4 shows an example of the meaning dictionary. The conversion candidate symbol storage unit 9 stores insertion target symbols for each symbol insertion position candidate.

  Note that the symbol input support device 10 may be configured not to include the symbol appearance frequency learning unit 3. In this case, learning is performed by another device, and the learning result is stored in the learning result storage unit 7. Further, the symbol input support device 10 may be configured not to include the character string conversion unit 5. In this case, for example, the processing result by the insertion symbol selection unit 4 and the input sentence are transmitted to another device including the character string conversion unit 5, and the symbol is inserted into the input sentence and output by the other device.

(Overview of operation)
The outline of the operation of the symbol input support device 10 having the configuration of FIG. 1 will be described below with reference to the flowchart shown in FIG.

  Step 11) First, the symbol appearance frequency learning unit 3 uses the text set and symbol list stored in the learning data storage unit 6 to generate co-occurrence word information that is co-occurrence information of symbols and words included in the text. Context information indicating what kind of context the symbol has appeared, co-occurrence information between the symbol and the word in the semantic dictionary given for each symbol, and the calculated result in the learning result storage unit 7 Store.

  Step 12) Next, the insertion symbol selection unit 4 uses the co-occurrence word information, context information, and co-occurrence information with words in the meaning dictionary at each symbol insertion position candidate in the input sentence, and learns for each symbol. Referring to the context information and co-occurrence information of the appearance word information stored in the result storage unit 7, the most appropriate symbol and insertion position in each symbol insertion position candidate are specified, and the processing result is stored in the conversion candidate symbol storage unit 9 To do.

  Step 13) The character string converter 5 inserts a symbol into the input sentence based on the symbol selected at step 12 and the insertion position, and outputs the sentence with the symbol inserted via the output unit 2.

(Details of operation of the symbol appearance frequency learning unit 3)
Next, the processing operation of the symbol appearance frequency learning unit 3 will be described in detail.
The symbol appearance frequency learning unit 3 performs learning using a text set including symbols in the symbol list stored in the learning data storage unit 6, so that the optimum position of the given input sentence is aligned with the context. A model for inserting symbols is generated as a learning result. In this learning, for each symbol of the symbol list appearing in the text set, co-occurrence information with the word, context information about what context the symbol appeared in, and the semantic dictionary given for each symbol and symbol The co-occurrence information with the word is calculated and the calculation result is stored in the learning result storage unit 7.

  In the present embodiment, the symbol appearance frequency learning unit 3 uses the Naive Bayes method as a learning method. Hereinafter, the Naive Bayes method will be described.

  A case x is expressed as an element vector as follows.

ｘの分類先のクラスを以下とする。

The class to which x is classified is as follows.

与えられた事例について、適した記号を決定するためには、P(c_i|x)を最大にするc_iを求めればよい。ベイズの定理から、

To determine a suitable symbol for a given case, we need to find c _i that maximizes P (c _i | x). From Bayes' theorem,

より、P(c_i)P(x|c_i)を最大にするc_iを求めればよい。ここで、

Thus, it is only necessary to obtain c _i that maximizes P (c _i ) P (x | c _i ). here,

と仮定してP(x|c_i)の計算を行うことにより、求めるc_iは、下記の式（５）と表される。

Assuming P with | by the calculation of (x c _i), it is c _i ask, denoted by the following equation (5).

本発明の実施の形態では、事例xを入力文、事例xにおける要素ベクトルの各ベクトルを、記号とほかの単語との共起情報、該記号がどのようなコンテクストによって出現したかというコンテクスト情報、該記号と記号ごとに与えられた語義辞書中の語との共起情報とし、分類先クラスを、事例xに挿入される各記号として、本手法に適用する。

In the embodiment of the present invention, case x is an input sentence, each vector of element vectors in case x is co-occurrence information of a symbol and another word, context information indicating what context the symbol appears in, The co-occurrence information between the symbol and the word in the semantic dictionary given for each symbol is used, and the classification destination class is applied to the present method as each symbol inserted into the case x.

In the learning performed by the symbol appearance frequency learning unit 3, P (x | c _i ) that is each vector element of the element vector is learned. Hereinafter, the process performed by the symbol appearance frequency learning unit 3 will be described with reference to the flowchart of FIG.

  Step 21) In step 21, a co-occurrence word importance calculation process is performed. That is, the symbol appearance frequency learning unit 3 reads the text set and the symbol list from the learning data storage unit 6, and for each symbol of the symbol list that appears in the text set, the symbol appearance frequency learning unit 3 shares all the words that co-occur with the symbol. The occurrence frequency is extracted, the co-occurrence relationship (corresponding to the importance) between the symbol and the word is calculated, and the calculation result is stored in the learning result storage unit 7. The calculation method of the co-occurrence relationship is as shown in the following formula (6).

Formula (6) P (word | symbol _i )
FIG. 7 shows examples of co-occurrence words and co-occurrence relationship calculation results. For example, in FIG. 7, the co-occurrence relationship between the symbol in the column whose ID is 1 and the word (send) is expressed as the probability (3/1000) as the calculation result of the above formula.

  Step 22) In step 22, context importance is calculated. Here, the symbol appearance frequency learning unit 3 extracts the co-occurrence frequency of the left and right context information (word, part of speech) of the appearance location for each symbol of the symbol list that appears in the text set, and the symbol and context And the calculated result is stored in the learning result storage unit 7. The calculation method of the co-occurrence relationship is as shown in the following formula (7).

Formula (7) P (context | symbol _i )
FIG. 8 shows an example of the calculation result of the co-occurrence relationship with the left and right words, and FIG. 9 shows an example of the calculation result of the co-occurrence relationship for each part of speech.

  Step 23) In step 23, the distance from the semantic dictionary is calculated. In other words, the symbol appearance frequency learning unit 3 refers to the meaning dictionary, so that each symbol of the symbol list that appears in the text set is co-occurrence with a word in the meaning dictionary (FIG. 4) given for each symbol. Information is calculated, and the calculation result is stored in the learning result storage unit 7. The calculation method of the co-occurrence relationship is as shown in the following formula (8).

Formula (8) P (word in dictionary | symbol _i )
(Details of operation of insertion symbol selection unit 4)
Next, the processing operation of the insertion symbol selection unit 4 will be described in detail.

  The inserted symbol selection unit 4 uses the co-occurrence word information, the context information, and the co-occurrence information with the words in the meaning dictionary in each symbol insertion position candidate in the input sentence, and the learning result stored in the learning result storage unit 7 Based on the above, in the symbol insertion position candidate, the most appropriate insertion candidate symbol among the symbols in the symbol list is specified. In the present embodiment, the symbol insertion position candidates are all morpheme parts and morpheme parts that are content words in the result of the morphological analysis of the input sentence. Hereinafter, the processing executed by the insertion symbol selection unit 4 will be described with reference to the flowchart of FIG.

  Step 31) The insertion symbol selection unit 4 performs morphological analysis of the input sentence input from the input unit 1.

  Step 32) Next, the inserted symbol selection unit 4 performs a symbol specifying process for each morpheme break. In other words, the insertion symbol selection unit 4 uses all morphemes analyzed in step 31 as symbol insertion position candidates, and co-occurrence with co-occurrence word information, context information, and words in the symbol dictionary in the symbol insertion position candidates. Based on the learning result stored in the learning result storage unit 7 using the information, the most appropriate symbol for insertion is specified from the symbols in the symbol list. FIG. 11 shows an example of symbol insertion position candidates in the input sentence. In FIG. 11, a portion indicated by an arrow is a symbol insertion position candidate.

  Step 33) Here, a symbol specifying process is performed for each content word morpheme. That is, the inserted symbol selection unit 4 uses the morpheme that is the content word among the morphemes analyzed in step 31 as the symbol insertion position candidate, and includes context information, co-occurrence word information, and symbol dictionary in the symbol insertion position candidate. Based on the learning result stored in the learning result storage unit 7, the most appropriate symbol for insertion is specified from the symbols in the symbol list. FIG. 12 shows an example of symbol insertion position candidates.

(Details of symbol identification processing for each morpheme break)
Hereinafter, the symbol specifying process (step 32 in FIG. 10) for each morpheme segment performed by the insertion symbol selection unit 4 will be described in more detail with reference to the flowchart in FIG.

  Here, all morphemes of the input sentence are set as symbol input position candidates, and the co-occurrence word information, the context information, and the co-occurrence information with words in the symbol dictionary in the symbol input position candidates are used, and the learning result storage unit 7 The most appropriate symbol for insertion and the insertion position among the symbols in the symbol list are specified based on the learning result stored in the table. In this way, by identifying all morpheme analyzed as morpheme candidates as symbol insertion position candidates, it is possible to specify a symbol along the end of a sentence or the way the symbols used between morphemes are used.

Step 41) The inserted symbol selection unit 4 extracts all morphemes in the input sentence analyzed in Step 31 of FIG. 10 as symbol input position candidates.
Step 42) The insertion symbol selection unit 4 performs co-occurrence when each symbol (c ₁ , c ₂ , c ₃ ,..., C _j ) in the symbol list is inserted into a specific symbol input position candidate. Word information, context information, and co-occurrence information with words in the symbol dictionary are calculated using the learning result stored in the learning result storage unit 7, and the sentence feature amount when the symbol is inserted into the input sentence is calculated. Calculate for each symbol. The calculation method is as shown in Equation (9), and the sentence feature amount is calculated as the appearance probability of the symbol. In the calculation of the sentence feature amount, when P (x _t | c _i ) that is each vector element of the element vector does not exist, 0 or a minute value is given.

ステップ４３）挿入記号選定部４は、前記の式（５）に基づいて、ステップ４２において記号一覧の記号ごとに与えられた文特徴量のうち、最も文特徴量が大きくなった記号（すなわち、確率が最も高い記号）を特定する。
ステップ４４）挿入記号選定部４は、ステップ４３で特定された記号の、該記号が入力文に挿入された場合の文特徴量が、予め定めた閾値以上であるか否かを確認する。

Step 43) The inserted symbol selection unit 4 determines the symbol having the largest sentence feature among the sentence features given for each symbol in the symbol list in Step 42 based on the above equation (5) (that is, The symbol with the highest probability).
Step 44) The inserted symbol selection unit 4 confirms whether or not the sentence feature amount of the symbol specified in Step 43 when the symbol is inserted into the input sentence is equal to or greater than a predetermined threshold value.

  If the sentence feature amount is equal to or greater than the threshold value as a result of the confirmation, the process proceeds to step 45, and if the sentence feature amount is less than the threshold value, the process proceeds to step 46.

  Step 45) It is assumed that the symbol set to be equal to or higher than the threshold value in Step 44 is a symbol to be inserted in the symbol input position candidate, and the symbol is specified.

  Step 46) It is determined that there is no symbol to be converted, assuming that there is no symbol to be inserted in the symbol input position candidate.

  Step 47) The inserted symbol selection unit 4 includes information indicating the current symbol input position candidate to be processed, the symbol and sentence feature specified in Step 45, or the information that there is no symbol specified in Step 46. Is stored in the conversion candidate symbol storage unit 9.

  Step 48) The insertion symbol selection unit 4 confirms whether or not the current symbol input position candidate is the last symbol input position candidate. If not, the process proceeds to the next symbol input position candidate. If it is the last, the process is terminated.

(Details of symbol identification for each content word morpheme)
Next, the symbol specifying process for each content word morpheme (step 33 in FIG. 10) executed by the insertion symbol selector 4 will be described in more detail with reference to the flowchart of FIG.

  Here, the morpheme portion of the content word of the input sentence is set as a symbol input position candidate, and the co-occurrence word information, context information, and co-occurrence information with the words in the symbol dictionary in the symbol input position candidate are used, and the learning result storage unit Based on the learning result stored in 7, the most appropriate symbol for insertion and the insertion position among the symbols in the symbol list are identified. In this way, by specifying a morpheme portion, which is a content word, among morphemes subjected to morpheme analysis as a symbol insertion position candidate, a symbol is identified in accordance with how the symbol used instead of the content word is used.

  Step 51) The inserted symbol selection unit 4 extracts morpheme portions of all content words of the input sentence subjected to morphological analysis in Step 31 of FIG. 10 as symbol input position candidates.

Step 52) The inserted symbol selection unit 4 inserts each symbol (c ₁ , c ₂ , c ₃ ,..., C _j ) in the symbol list into each symbol input position candidate extracted in Step 51. When the co-occurrence word information, context information, and co-occurrence information with the words in the symbol dictionary are calculated using the learning result stored in the learning result storage unit 7, and the symbol is inserted into the input sentence The sentence feature amount of is calculated. Here, in order to identify the symbols along the way the symbols used instead of the content words are used, the target word to be browsed when calculating the co-occurrence relationship with the words in the symbol dictionary is the symbol input position candidate The content word is The sentence feature amount calculation method is as shown in Equation (9), and the sentence feature amount is calculated as the appearance probability of the symbol. In the calculation of the sentence feature amount, when P (x _t | c _i ) that is each vector element of the element vector does not exist, 0 or a minute value is given.

Step 53 to step 58 are the same as step 43 to step 48.
(Details of operation of the character string converter 5)
Next, the processing operation of the character string converter 5 will be described in detail.

  The character string conversion unit 5 performs processing for inserting the insertion target symbol in each symbol insertion position candidate stored in the conversion candidate symbol storage unit 9 into the input sentence. Here, when insertion is performed according to the result stored in the conversion candidate symbol storage unit 9, a plurality of the same symbols may appear in the input sentence. Using the given sentence feature, a plurality of the same symbols are prevented from appearing in the input sentence. Hereinafter, the processing executed by the character string conversion unit 5 will be described with reference to the flowchart of FIG.

  Step 61) The character string conversion unit 5 extracts the symbols of the respective symbol insertion position candidates from the conversion candidate symbol storage unit 9.

  Step 62) The character string conversion unit 5 checks whether or not the symbols extracted from the conversion candidate symbol storage unit 9 include a plurality of identical symbols. If included, the process proceeds to step 63. If not included, the process proceeds to step 65 where the symbol is stored in the conversion candidate symbol storage unit 9.

  Step 63) The character string conversion unit 5 selects a symbol more suitable for the input sentence from among the plurality of the same symbols included in the conversion candidate symbol storage unit 9, so that the sentences given to the plurality of the same symbols respectively Of the feature quantities, a symbol having the largest sentence feature quantity is set as a conversion candidate, and the conversion candidate symbol storage unit 9 stores the symbol. Of the plurality of identical symbols, symbols other than the symbol having the largest sentence feature value are not stored in the conversion candidate symbol storage unit 9. Such processing is performed for each of a plurality of sets of identical symbols.

  Step 64) The character string conversion unit 5 inserts the symbol stored in the conversion candidate symbol storage unit 9 by the above processing at each insertion position in the input sentence, and outputs the sentence into which the symbol is inserted as the output result. Output via.

  The symbol input support device 10 described so far can be realized, for example, by causing a computer having a CPU, a memory, and the like to execute a program corresponding to each process. The program may be distributed by being recorded on a computer-readable recording medium such as a portable memory, or may be downloaded from a server on the network. As described above, when the symbol input support device 10 is realized by a computer, processing data such as a character string is stored in a memory or the like, and is appropriately read and processed by the CPU in accordance with an instruction of the program. The “computer” is not limited to a PC terminal, and is used to include various devices having functions equivalent to those of a computer, such as a mobile phone.

  Further, the symbol input support device 10 may be realized by a single computer or may be realized by connecting a plurality of computers via a network. For example, configuring one or more of the plurality of storage units as an external server device is an example when the symbol input support device 10 is realized by a plurality of computers. In addition, the symbol appearance frequency learning unit 3 and the other functional units may be configured by separate computers as an example when the symbol input support device 10 is realized by a plurality of computers. Furthermore, the character string conversion unit 5 and other functional units may be configured by separate computers.

(Effect of embodiment)
According to the technology according to the present embodiment, when a symbol is specified, the relationship between the symbol and the sentence feature amount, the relationship between the symbol and the appearance position, and the relationship between the symbol and the role are learned and learned in advance. Since the insertion is performed based on the result, it is possible to obtain a result reflecting the actual usage of the symbol in consideration of the content of the sentence.

  In addition, when learning is performed at the appearance of a symbol, it is possible to insert the symbol at an appearance position along the actual usage by performing learning focusing on the appearance position of the symbol and the symbol.

  In addition, by learning co-occurrence word information, context information, and co-occurrence information with words in the symbol dictionary of sentences containing symbols, it is possible to insert symbols according to the context. Become.

  Furthermore, by performing learning focusing on the role that the symbol is used as a word, the insertion considering the role of the symbol can be performed.

The present invention can be applied to a device that inserts a symbol suitable for the context of an input sentence at an appropriate position of the input sentence.
The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

DESCRIPTION OF SYMBOLS 1 Input part 2 Output part 3 Symbol appearance frequency learning part 4 Inserted symbol selection part 5 Character string conversion part 6 Learning data storage part 7 Learning result storage part 8 Meaning dictionary storage part 9 Conversion candidate symbol storage part 10 Symbol input assistance apparatus

Claims (9)

A symbol input support device for inserting a symbol into an input sentence,
For each symbol in the learning sentence, co-occurrence word information indicating the frequency of appearance with the co-occurring word, context information, and the co-occurrence information of the symbol and the word in the predetermined dictionary given for each symbol, Symbol appearance frequency learning means for storing the calculation result as a learning result in the learning result storage means,
For each symbol insertion position candidate in the input sentence, a symbol is inserted at each position of the symbol insertion position candidate using the learning result stored in the learning result storage unit for each symbol to be inserted. An insertion symbol selection unit that calculates a sentence feature amount in a case, identifies a symbol and an insertion position most suitable for insertion based on the sentence feature amount, and stores the symbol and the insertion position in a conversion candidate symbol storage unit;
A symbol input support device comprising:   The inserted symbol selection means includes co-occurrence word information, context information, and co-occurrence information with words in the predetermined dictionary when a symbol is inserted at the position of the symbol insertion position candidate in the input sentence. 2. The symbol input support device according to claim 1, wherein the symbol input support device calculates the symbol appearance probability based on the learning result as the sentence feature amount.   3. The symbol input support according to claim 1, further comprising character string conversion means for inserting a symbol into the input sentence using the symbol specified by the insertion symbol selecting means and the insertion position. apparatus.   When the plurality of identical symbols are specified by the insertion symbol selecting means, the character string converting means is configured to use the plurality of identical symbols based on the sentence feature amounts corresponding to the plurality of identical symbols. The symbol input support device according to claim 3, wherein one of the symbols is selected and the symbol is inserted at an insertion position corresponding to the symbol in the input sentence. A symbol input support method executed by a symbol input support device for inserting a symbol into an input sentence,
For each symbol in the learning sentence, co-occurrence word information indicating the frequency of appearance with the co-occurring word, context information, and the co-occurrence information of the symbol and the word in the predetermined dictionary given for each symbol, And a symbol appearance frequency learning step of storing the calculation result as a learning result in the learning result storage means;
For each symbol insertion position candidate in the input sentence, a symbol is inserted at each position of the symbol insertion position candidate using the learning result stored in the learning result storage unit for each symbol to be inserted. An insertion symbol selection step of calculating a sentence feature amount in the case, identifying a symbol and insertion position most suitable for insertion based on the sentence feature amount, and storing the symbol and insertion position in the conversion candidate symbol storage means;
A symbol input support method characterized by comprising:   In the inserted symbol selection step, the symbol input support device includes co-occurrence word information, context information, and words in the predetermined dictionary when a symbol is inserted at the position of the symbol insertion position candidate in the input sentence. The symbol input support method according to claim 5, wherein the co-occurrence information is obtained using the learning result, and a symbol appearance probability based on the information is calculated as the sentence feature amount.   The symbol input support according to claim 5 or 6, further comprising a character string conversion step of inserting a symbol into the input sentence using the symbol specified by the insertion symbol selection step and the insertion position. Method.   In the character string conversion step, the symbol input support device, when a plurality of identical symbols are specified by the insertion symbol selection step, based on respective sentence feature amounts corresponding to the plurality of identical symbols. The symbol input support method according to claim 7, wherein one of the plurality of identical symbols is selected, and the symbol is inserted at an insertion position corresponding to the symbol in the input sentence. .   The program for making a computer perform each step in the symbol input assistance method of any one of Claim 5 thru | or 8.

Images