JP5510263B2 - Emoticon reading information estimation device, emoticon reading information estimation method, emoticon reading information estimation program, and information terminal - Google Patents

Description

  The present invention relates to an emoticon reading information estimation device, an emoticon reading information estimation method, an emoticon reading information estimation program, and an information terminal.

  In the field of speech synthesis, methods to examine and develop emoticons (“(^ _ ^)”, “(T_T)”, etc.) represented by combinations of multiple symbols as meanings represented by emoticons are being studied and developed. (For example, see Patent Documents 1 to 3). In these methods, emoticons (“(^ _ ^)”, “(T_T)”, etc.) are not read out as symbols like “braces, hats, underbars…”, but like “Nikori” or “Aen” To read aloud. In the present specification, for convenience of explanation, a character string enclosed in double brackets “” means an emoticon or a component of the emoticon. That is, the double angle bracket “” does not constitute part of the emoticon.

  In the above-mentioned Patent Document 1, there is a technique in which a plurality of emoticon expressions and corresponding readings are registered in a dictionary in advance and the meanings represented by the emoticons are read out by searching for emoticon readings from the dictionary. It is disclosed. Moreover, in patent document 2, the character used as the outline of an emoticon is defined beforehand as an outline symbol table, the emoticon part is detected using the outline symbol, and according to the number of characteristic characters existing inside the outline. A technique for determining how to read is disclosed. Furthermore, in Patent Document 3, characters (symbols) and meanings are defined in advance for each facial part, such as an eye part and a mouth part, and the meaning represented by the whole emoticon is analyzed by stacking the parts. Technology is disclosed.

JP-A-11-305987 JP 2002-268665 A JP 2005-284192 A

  However, in Patent Document 1, only emoticons registered in the dictionary in advance can be read out, and emoticons rich in variations cannot be read out. Furthermore, in Patent Document 2, emoticons that do not have all or part of outline symbols (such as “^ _ ^;”, “´ω`) ノ”) and emoticons that use contour symbols in a different meaning are used. (For example, "(') Д (｀)" whose outline symbol (parentheses) represents a cheek) cannot be read out.

  Also, in Patent Documents 2 and 3, when the same symbol is used as different facial parts (such as “(゜ ∋ ゜)” or “∈ (゜ ◎ ゜) 同 じ” where the same symbol is used as a hand and nose) If the expected facial parts are not available (such as “(´Д⊂ グ ス ン”), where one eye does not exist, it cannot be read out correctly.

  Therefore, the present case has been made in view of the above problems, and for emoticons reading information estimation device, emoticon reading information capable of accurately estimating reading information representing the meaning of various emoticons. An object is to provide an estimation method and an emoticon reading information estimation program. In addition, an object of the present invention is to provide an information terminal that can provide an appropriate reading of an emoticon to a user.

  The emoticon reading information estimation apparatus described in this specification is an emoticon reading information estimation apparatus that estimates reading information of an emoticon represented by a plurality of symbols. A storage unit for storing an emoticon dictionary in which at least one main symbol extracted in advance as a contribution to the meaning content of the emoticon and the reading information of the emoticon are associated and registered; For each emoticon registered in the emoticon dictionary and a reception unit that accepts an input of the emoticon from the main symbol and a plurality of symbols representing the emoticon input to the reception unit, An acquisition unit that acquires a ratio of the same symbol as the main symbol in all of a plurality of symbols that represent the input emoticon, the input emoticon based on the ratio acquired by the acquisition unit, Registered in the emoticon dictionary A calculation unit for calculating the similarity to the emoticon to be detected, and based on the similarity, the reading information of the inputted emoticon is estimated from the emoticon reading information registered in the emoticon dictionary An emoticon reading information estimation device comprising:

  The information terminal described in this specification includes the emoticon reading information estimation device described in this specification and the input emoticon reading information estimated by the estimation unit of the emoticon reading information estimation device. And an output device that outputs as character data or voice.

  The emoticon reading information estimation method described in the present specification is an emoticon reading information estimation method for estimating emoticon reading information represented by a plurality of symbols using a computer, and inputs emoticons from the outside A reception step of accepting a message, at least one main symbol extracted in advance as a contribution to the meaning content of the emoticon from a plurality of symbols representing the emoticon, and the reading information of the emoticon With reference to the emoticon dictionary that is attached and registered, for each emoticon registered in the emoticon dictionary, the main symbol is compared with a plurality of symbols representing the input emoticon, An acquisition step of acquiring a ratio of the same symbol as the main symbol in a plurality of symbols representing the input emoticon, the input emoticon based on the ratio acquired in the acquisition step, and the face In the character dictionary A calculation step of calculating a similarity to the recorded emoticon, and reading of the input emoticon from the emoticon reading information registered in the emoticon dictionary based on the similarity An estimation method for estimating information, and a emoticon reading information estimation method executed by a computer.

  The emoticon reading information estimation program described in this specification is an emoticon reading information estimation program that causes a computer to estimate reading information of an emoticon represented by a plurality of symbols. The reception process of accepting the input of at least one main symbol extracted in advance as a contribution to the semantic content of the emoticon from a plurality of symbols representing the emoticon, and the reading information of the emoticon With reference to the emoticon dictionary registered in association with each other, for each emoticon registered in the emoticon dictionary, the main symbol is compared with a plurality of symbols representing the input emoticon, An acquisition step of acquiring a ratio of the same symbol as the main symbol in all of a plurality of symbols representing the input emoticon, and the input face sentence based on the ratio acquired in the acquisition step And a calculation step of calculating a similarity with the emoticon registered in the emoticon dictionary, and the reading information of the emoticon registered in the emoticon dictionary based on the similarity, An emoticon reading information estimation program that executes an estimation step of estimating input reading information of an emoticon.

  The emoticon reading information estimation device, emoticon reading information estimation method, and emoticon reading information estimation program described in this specification accurately estimate reading information representing the meaning of various types of emoticons. There is an effect that is possible. In addition, the information terminal described in the present specification has an effect that it is possible to provide an appropriate reading of the emoticon to the user.

It is a block diagram which shows roughly the structure of the mobile telephone which concerns on one Embodiment. It is a figure which shows the hardware constitutions of the process part of FIG. It is a functional block diagram of a processing part. It is a block diagram of the emoticon extraction part of FIG. It is a block diagram of the emoticon reading estimation part of FIG. It is a flowchart which shows a series of processes of the mobile telephone (processing part) which concerns on one Embodiment. It is a flowchart which shows the specific processing content in step S3 of FIG. FIGS. 8A to 8D are diagrams (part 1) for explaining the processing of FIG. FIG. 9A to FIG. 9D are diagrams (part 2) for explaining the processing of FIG. It is a flowchart which shows the specific processing content in step S5 of FIG. It is a figure which shows the specific registration content of an emoticon dictionary. It is a figure which shows the matching result output to the similarity calculation part. It is a flowchart which shows the specific processing content in step S70 of FIG. It is a figure for demonstrating the process of FIG. It is a flowchart which shows the specific processing content in step S71 of FIG. It is a figure for demonstrating the process of FIG. It is a figure which shows a dictionary component order and an unknown face character matching order. It is a flowchart which shows the specific processing content in step S72 of FIG. It is FIG. (1) for demonstrating the process of FIG. FIG. 19 is a second diagram for explaining the processing in FIG. 18; FIG. 19 is a diagram (No. 3) for explaining the processing of FIG. 18; It is FIG. (4) for demonstrating the process of FIG.

  Hereinafter, an embodiment will be described in detail with reference to FIGS. FIG. 1 is a block diagram showing a configuration of a mobile phone 100 according to an embodiment. As shown in FIG. 1, the mobile phone 100 includes an input unit 52, a display display 54, a speaker 56, and a processing unit 50.

  The input unit 52 includes a keyboard and a touch panel, and is an interface for a user to input an instruction to the processing unit 50. In addition, the input unit 52 includes a microphone used for a call and the like, and a communication device that receives text transmitted by wireless communication such as mail and website data. The display 54 includes a liquid crystal display, an organic EL display, and the like, and displays a processing result by the processing unit 50. In addition to outputting the voice of the other party, the speaker 56 outputs a reading voice of a mail or a document in the browser by the reading function of the processing unit 50.

  The processing unit 50 performs various processes (call, mail, Internet, camera photography, etc.) in a general mobile phone, and executes a document reading process. FIG. 2 shows a hardware configuration of the processing unit 50. As shown in FIG. 2, the processing unit 50 includes a CPU 90, a ROM 92, a RAM 94, a storage unit (in this example, an HDD (Hard Disk Drive)) 96, an input / output unit 97, and the like. Each component of the processing unit 50 is connected to the bus 98. In the processing unit 50, the CPU 90 executes a program (including an emoticon reading information estimation program) stored in the ROM 92 or the HDD 96, thereby realizing the functions of the respective units in FIG. The input / output unit 97 is connected to the input unit 52, the display 54, the speaker 56, and the like shown in FIG. In the above description, an HDD is used as the storage unit 96. However, the present invention is not limited to this, and an SSD (Solid State Drive) may be used from the viewpoint of impact resistance and power saving.

  FIG. 3 is a functional block diagram showing functions related to reading a document in the processing unit 50. As shown in FIG. 3, the processing unit 50 functions as a language analysis unit 2 as an emoticon reading information estimation device and a waveform processing unit 3 as an output device. The language analysis unit 2 analyzes the input text and emoticon and estimates the reading and accent. The waveform processing unit 3 generates a speech waveform from the reading information analyzed by the language analysis unit 2.

  More specifically, as shown in FIG. 3, the language analysis unit 2 is a text input receiving unit 21, a face character extraction unit 22, a language dictionary storage unit 23, a normal text analysis unit 24, and a face character dictionary storage as a storage unit. Unit 25, emoticon reading estimation unit 26, and reading output unit 27.

  The text input accepting unit 21 accepts input of text information and outputs a text character string to the emoticon extracting unit 22. In addition, as a text input method to the text input reception part 21, the direct input method from a keyboard and the input method by communication via a wired or wireless network are employable. The text input to the text input receiving unit 21 may be acquired from a recording medium such as a CD (Compact Disc) or a DVD (Digital Versatile Disc), or image data scanned using a scanner is OCR (Optical Character Reader) may be processed and acquired. Furthermore, other general text input methods may be adopted as appropriate.

  The emoticon extraction unit 22 receives the text character string output from the text input reception unit 21, and extracts a portion where two or more symbols are continuous as an emoticon. The extracted emoticon is output to the emoticon reading estimation unit 26, and the normal text that is not the emoticon is output to the normal text analysis unit 24.

  Here, details of the emoticon extraction unit 22 will be described with reference to FIG. The emoticon extraction unit 22 includes an input buffer 221, a character determination unit 224, an emoticon buffer 222, a normal character string buffer 223, and an output unit 225. The input buffer 221 stores the input text output from the text input receiving unit 21. The character determination unit 224 determines that a portion in which two or more symbols are continuous in the input text is a face character, and stores the portion in the face character buffer 222. Further, the character determination unit 224 determines that the portion other than the emoticon is a normal character string and stores it in the normal character string buffer 223. The emoticon buffer 222 stores emoticons extracted from the input text, and the ordinary character string buffer 223 stores ordinary character strings other than the emoticons in the input text. The output unit 225 outputs the character string stored in the emoticon buffer 222 to the emoticon reading estimation unit 26, and outputs the character string stored in the normal character string buffer 223 to the normal text analysis unit 24. To do.

  Returning to FIG. 3, the normal text analysis unit 24 refers to the language dictionary storage unit 23 to determine the reading and accent of the text, and outputs the reading and other prosodic symbols to the reading output unit 27. Here, generally known morphological analysis processing can be used as the text analysis processing. As an output format, the Kana level notation of the Japanese text-to-speech symbol shown in JEITA IT-4002 can be used.

  The emoticon dictionary storage unit 25 stores an emoticon dictionary as shown in FIG. As shown in FIG. 11, in the emoticon dictionary, “notation”, “reading / accent” as reading information, “main part”, and “subordinate part” are serial numbers (data No.) for each emoticon. .) Is associated and registered.

  In “notation”, symbols of all known emoticons are registered. In “Reading / Accent”, a reading and an accent representing the meaning of the emoticon are registered. Readings and accents can be registered by, for example, Kana level notation of Japanese text-to-speech synthesis symbols shown in JEITA IT-4002.

  The “main part” is a main symbol extracted in advance as being responsible for the meaning of the emoticon, i.e., contributing to the meaning content of the emoticon, for example, overlapping between a plurality of known emoticons representing the same meaning. It is a symbol. The “subordinate part” is a symbol having a thin relationship with the meaning of the emoticon and is a subordinate symbol other than the main part. Note that only one consecutive identical symbol may be registered.

  For example, suppose that there are three known emoticons with the meaning of “Wai”: “\ (^ O ^) /” “\ (^^ @) /” “\ ^ o ^ /”. Symbols common to these are “\”, “^” on the left side, “^” on the right side, and “/”. Therefore, these four symbols are the main parts of “Wai”. Then, a part obtained by removing main parts from each emoticon becomes a subordinate part of each emoticon. That is, if the emoticon is “\ (^ O ^) /”, the main part is “\ ^^ /” and the subordinate part is “(O)”. If the emoticon is “\ (^^ @) /”, the main part is “\ ^^ /” and the subordinate part is “(@)”. Furthermore, if the emoticon is “\ ^ o ^ /”, the main part is “\ ^^ /” and the subordinate part is “o”.

  As a method for extracting main parts, a part common to a certain number of emoticons among a plurality of known emoticons representing the same meaning may be used as the main part. By doing so, it becomes possible to deal with more variations.

  Returning to FIG. 3, the emoticon reading estimation unit 26 calculates the similarity between the input emoticon and the emoticon registered in the emoticon dictionary of FIG. 11, and based on the similarity, Estimate the reading of the input emoticon. However, if the similarity does not exceed a predetermined threshold value, it is determined that the character is not an emoticon and is output to the normal text analysis unit 24. The normal text analysis unit 24 analyzes the reading as a symbol.

  Here, as shown in FIG. 5, the emoticon reading estimation unit 26 includes an emoticon input reception unit 260 as a reception unit, a ratio acquisition unit 261 as an acquisition unit, and a similarity calculation unit 262 as a calculation unit. And a reading estimation unit 264 as an estimation unit. The emoticon input accepting unit 260 accepts an input of the emoticon output from the output unit 225 of the emoticon extracting unit 22. The ratio acquisition unit 261 includes a main component associating unit 2611 and a dependent component associating unit 2612. Based on the data output from the ratio acquisition unit 261, the similarity calculation unit 262 calculates the similarity between the emoticon input from the emoticon extraction unit 22 and each emoticon registered in the emoticon dictionary. To do. The reading estimation unit 264 estimates the reading of the emoticon input from the emoticon extraction unit 22 based on the similarity calculated by the similarity calculation unit 262, and outputs it to the reading output unit 27. However, if the similarity does not exceed a predetermined threshold value, it is determined that the character is not an emoticon, and the text received by the emoticon input receiving unit is output to the normal text analyzing unit 24.

  The reading output unit 27 sequentially joins the reading of the normal text output from the normal text analysis unit 24 and the reading of the emoticon output from the emoticon reading estimation unit 26 to obtain a final reading as a waveform processing unit. Output to 3.

  The waveform processing unit 3 includes a prosody dictionary storage unit 31, a waveform dictionary storage unit 32, and a voice generation unit 33. The prosodic dictionary storage unit 31 stores a prosodic dictionary in which data representing characteristics of how to speak humans such as intonation and rhythm are registered. The waveform dictionary storage unit 32 stores a waveform dictionary in which human voice data itself or a voice feature amount extracted from voice data is registered. The voice generation unit 33 generates a voice from the text reading input from the reading output unit 27 and the data of the prosodic dictionary and the waveform dictionary. Note that a known method such as a corpus-based speech synthesis method or a formant synthesis method can be employed as a method for generating the speech by the speech generation unit 33.

  Next, the text reading process in the processing unit 50 will be described in detail with reference to FIGS. First, in step S1 of FIG. 6, the text input reception part 21 acquires an input text using the various methods mentioned above. Next, in step S <b> 2, the emoticon extraction unit 22 takes in the input text output from the text input reception unit 21.

  Next, in the subroutine of step S3, the emoticon extraction unit 22 executes a process of extracting an analysis unit from the input text. The analysis unit means a continuous section of a character string of two or more characters or a continuous section of other character strings. In this step S3, the emoticon extraction unit 22 executes the process according to FIG. In the description of the processing in FIG. 7, as an example, it is assumed that the emoticon extraction unit 22 takes in the input text “sad (′ ∩ ｀)” shown in FIG.

  In the process of FIG. 7, first, in step S21, the character determination unit 224 sets a pointer at the head of the input buffer 221 (see FIG. 4). Specifically, as shown in FIG. 8A, a pointer is set at the character “sad” in the input text “sad ('∩ ｀.)”.

  Next, in step S22, the character determination unit 224 determines whether or not the character pointed to by the pointer and the character immediately following it are symbols. Here, since the character “sad” pointed to by the pointer and the character “shi” immediately after it are not symbols, the determination in step S22 is denied and the process proceeds to step S27.

  In step S27, the character determination unit 224 deletes the character indicated by the pointer from the input buffer 221, and stores the deleted character in the normal character string buffer 223. Next, in step S28, the character determination unit 224 advances the pointer of the input buffer 221 by one character. In the next step S29, the character determination unit 224 determines whether the pointer exceeds the last character of the input buffer 221, or whether the character pointed to by the pointer and the character immediately after the character are both symbols. When judgment here is denied, it returns to step S22. In the example of FIG. 8A, as shown in FIG. 8B, until the three characters “sad” are stored in the normal character string buffer 223, the processes and determinations in steps S22 and S27 to S29 are performed. Repeated. Then, as shown in FIG. 8B, when the character pointed to by the pointer is the symbol “(”, and the character immediately after it is “′”, the determination in step S29 is affirmed, and the process proceeds to step S30. Transition.

  In step S30, the output unit 225 outputs the contents of the normal character string buffer 223 to the normal text analysis unit 24 of FIG. 3 (see FIG. 8C). Then, in the next step S31, the character determination unit 224 left-justifies the data in the input buffer 221 as much as the deleted characters in the input buffer 221 as shown in FIG. Thereafter, the process proceeds to step S4 in FIG.

  In step S4 in FIG. 6, the emoticon extraction unit 22 determines whether or not the output character string is an emoticon. Here, since the output unit 225 outputs the content “sad” in the normal character string buffer 223 to the normal text analysis unit 24 in FIG. 3, the determination in step S4 is denied and the process proceeds to step S6. In step S <b> 6, the normal text analysis unit 24 analyzes the text reading based on the language dictionary stored in the language dictionary storage unit 23, and outputs the analyzed reading information to the reading output unit 27. Next, in step S <b> 7, the reading output unit 27 sequentially stores reading information in a buffer held by the reading output unit 27. Next, in step S8, the character determination unit 224 of the emoticon extraction unit 22 determines whether or not the input buffer 221 is empty. Here, since the text “(′ (.)” Still remains in the input buffer 221, the determination is denied and the process returns to step S3 (step S21 in FIG. 7).

  In step S21 of FIG. 7, the character determination unit 224 sets the pointer to the head of the input buffer 221 as shown in FIG. In step S22, the character determination unit 224 determines whether the character pointed to by the pointer and the character immediately after the character are symbols. Here, since the character “(” pointed to by the pointer and the character “′” immediately after it are symbols, the determination is affirmed and the process proceeds to step S23.

  In step S <b> 23, the character determination unit 224 deletes the character pointed to by the pointer from the input buffer 221 and stores the deleted character in the emoticon buffer 222. Next, in step S24, the character determination unit 224 advances the pointer of the input buffer 221 by one character. In the next step S25, the character determination unit 224 determines whether the pointer exceeds the last character of the input buffer or whether the character pointed to by the pointer is other than a symbol. If the determination here is negative, the process returns to step S23. Thereafter, steps S23 to S25 are repeated, and when the pointer exceeds the last character of the input buffer (see FIG. 9B), the process proceeds to step S26.

  In step S26, the output unit 225 outputs the contents of the emoticon buffer 222 to the emoticon reading estimation unit 26 (see FIG. 9C). Next, in step S31, the character determination unit 224 left-justifies the data in the input buffer 221 corresponding to the deleted input buffer. In FIG. 9D, no data exists in the input buffer 221. Thereafter, the process proceeds to step S4 in FIG.

  In step S4 in FIG. 6, the emoticon extraction unit 22 determines whether or not the output character string is an emoticon. If the determination here is affirmed, the process proceeds to step S5. Here, since the output unit 225 outputs the contents “(′ ∩ ｀.)” Of the normal character string buffer 222 to the emoticon reading estimation unit 26, the determination in step S4 is affirmed and the process proceeds to step S5. In step S5, the emoticon reading estimation unit 26 executes a subroutine of emoticon reading estimation processing. Specifically, in step S5, the emoticon reading estimation unit 26 executes the process of FIG.

  Here, as described above, the emoticon input reception unit 260 of the emoticon reading estimation unit 26 is input with the emoticon “(′ ∩ ｀.)” Shown in FIG. To do. It is assumed that this emoticon reads “Gusun”. However, in the emoticon dictionary shown in FIG. 11, the same notation as “(′ ∩ ｀.)” Is not registered, so the input emoticon is “unknown emoticon”.

  In FIG. 10, first, in step S <b> 61, the ratio acquisition unit 261 creates a variable N representing a data number and sets 1 as an initial value of the variable. Next, in step S62, the ratio acquisition unit 261 acquires the Nth (here, the first) data from the emoticon dictionary (see FIG. 11). In this case, the ratio acquisition unit 261 includes data No. 1 in FIG. Get the data with 1 (data with the notation "(^-^)").

  Next, in step S63, the main part associating unit 2611 sets a part number for each main part “^^” of the extracted data No. 1. It should be noted that as the part number of the main part, a character string that is a combination of the initial M and the serial number is set. In the case of the main part “^^”, the main part associating unit 2611 sets M1 in “^” on the left side and M2 in “^” on the right side.

  Next, in step S64, the main part association unit 2611 associates whether or not there is a symbol that matches the main part extracted from the emoticon dictionary in the input emoticon “(′ ∩ ｀.)”. I do. Then, the main component correlation unit 2611 outputs the correlation result (matching component number) to the similarity calculation unit 262 together with the number of main components in the emoticon dictionary. Here, M1 = “^”, M2 = “^”, and the number of main parts is 2. However, since the parts that match M1 and M2 are not included in the unknown emoticons, the main parts association unit 2611 outputs a null character as the association result (refer to the data of the cell where the row of “Data No.” 1 in FIG. 12, the column of “main component association result”, and the column of “number of dictionary main components” intersect). . When there is a matching that matches, the main component correlation unit 2611 outputs the matching component number and component notation to the similarity calculation unit 262 as the correlation result.

  Next, in step S65, the dependent component associating unit 2612 sets a component number for the dependent component "(-)" of the extracted data No. 1. As the part number of the dependent part, a character string of a combination of a subsidiary initial S and a serial number is set. Here, the dependent component associating unit 2612 sets the first component “(” as S1, the middle component “-” as S2, and the last component “)” as S3.

  Next, in step S66, the dependent component associating unit 2612 determines whether a symbol that matches the dependent component extracted from the emoticon dictionary exists in the input unknown emoticon “(′ ∩ ｀.)”. Perform the association. Then, the dependent component associating unit 2612 outputs the matching component number and component notation to the similarity calculating unit 262 as a result of the association. Here, S1 = “(), s2 =“ − ”, s3 =“) ”, and S1 and S3 match the input parts of the unknown emoticon, so the matching result is S1 (and S3) is output (refer to the data of the cell where the row of “data No.” 1 in FIG. 12 and the column of “subordinate component correspondence result” intersect).

  Next, in step S67, the ratio acquisition unit 261 increases the data number N by 1 (N ← N + 1). In step S68, it is determined whether N exceeds the number of data in the emoticon dictionary. If the determination here is negative, the process returns to step S62, and the same processing as described above is performed for the data No. of the emoticon dictionary. Execute for emoticons of = N. Thereafter, when the determination in step S68 is affirmed, the process proceeds to step S69A. Note that the association result output to the similarity calculation unit 262 at the stage of proceeding to step S69A is data as shown in FIG.

  In step S69A, the similarity calculation unit 262 calculates a temporary similarity. In this case, the similarity calculation unit 262 calculates, as a temporary similarity, a value obtained by dividing the number of components associated as main components by the number of main components in the emoticon dictionary for all data in FIG. . In step S <b> 69 </ b> B, the similarity calculation unit 262 extracts a data number that is greater than or equal to a certain threshold (first threshold). For example, in the case of the association result shown in FIG. Since the number of parts 2 associated with the main part is 0, the provisional similarity is 0%. On the other hand, since data No. 3 has a main component matching result of 3 and the number of dictionary main components is 3, the provisional similarity is 100% (= 3/3 × 100). Furthermore, since data No. 4 has a main part correspondence result of 3 and a dictionary main part number of 5, the temporary similarity is 60% (= 3/5 × 100). Here, for example, if the threshold value (first threshold value) used in step S69B is “60%”, data No. 3 and data No. 4 having a provisional similarity equal to or greater than the threshold value are represented by the similarity. The calculation unit 262 extracts the data.

  Next, in step S69C, the similarity calculation unit 262 creates a variable K representing the emoticon number, and sets 1 as the initial value of the variable. Next, in step S70, the similarity calculation unit 262 obtains a K-th (here, first) emoticon (here, emoticon of data No. 3) part number sequence among the extracted emoticons. Execute. In step S70, specifically, the similarity calculation unit 262 executes processing according to the flowchart of FIG.

  In the process of FIG. 13, first, in step S7001, the similarity calculation unit 262 sets a dictionary notation pointer at the beginning of the dictionary notation (hereinafter, pointers indicating various data in the dictionary are directly connected to the emoticon dictionary storage unit 25. May be pointed to the corresponding position on the RAM or cache after the data in the emoticon dictionary storage unit 25 is once read into the RAM or cache. Use a method that is convenient for implementation). Specifically, the similarity calculation unit 262 aligns the dictionary notation pointer with the “(” part of the emoticon “(′ _ ｀.)” (See the first step in FIG. 14). In S7002, the similarity calculation unit 262 sets the main component pointer to the head of the main component, creates a variable for storing the main component number, and sets it to 1. Specifically, the similarity calculation unit 262 The main component pointer is set to “′” in the main component “′ ｀.” (See the first stage in FIG. 14) Next, in step S7003, the similarity calculation unit 262 subordinates the subordinate component pointer. Set at the beginning of the part, create a variable to store the dependent part number, and set it to 1. More specifically, the similarity calculation unit 262 sets the dependent part pointer to “(())” among the dependent parts “(_)”. ] (See the first stage of the first time in FIG. 14) .

  Next, in step S7004, the similarity calculation unit 262 determines whether or not the symbols indicated by the dictionary notation pointer and the main component pointer match. Here, as shown in the first row of the first time in FIG. 14, the two do not match, so the determination is negative, and the flow proceeds to step S7007. In step S7007, the similarity calculation unit 262 determines whether or not the symbols indicated by the dictionary notation pointer and the subordinate component pointer match. Here, as shown in the first stage of FIG. 14, since they are the same, the determination is affirmed and the process proceeds to step S7008.

  In step S7008, the similarity calculation unit 262 outputs the initial S representing the dependent component and the dependent component number (S1) to the output buffer of the similarity calculation unit 262. In the next step S7009, the similarity calculation unit 262 advances the dependent component pointer to the next symbol, and adds 1 to the dependent component number (dependent component number = 2). The results of steps S7008 and S7009 are shown in the first second row in FIG.

  Next, in step S7010, the similarity calculation unit 262 advances the dictionary notation pointer to the next symbol. This result is shown in the first third row of FIG.

  Next, in step S7011, the similarity calculation unit 262 determines whether or not the dictionary notation pointer points to a symbol. Here, since the dictionary notation pointer points to the symbol, the determination is denied, and the process returns to step S7004.

  Next, the second repetition will be described. In the second repetition, as shown in the first row of the second time in FIG. 14, the symbol “′” pointed to by the dictionary notation pointer and the main part pointer matches, so the determination in step S7004 is affirmed, and step S7005 Migrate to In step S7005, the similarity calculation unit 262 outputs the initial value M representing the main part and the main part number (M1) included in the similarity calculation unit 262, as shown in the second row in FIG. Output to buffer.

  Next, in step S7006, the similarity calculation unit 262 advances the main part pointer to the next symbol “｀” and adds 1 to the main part number as shown in the third row of the second time in FIG. Part number = 2). Next, in step S7010, the similarity calculation unit 262 advances the dictionary notation pointer to the next symbol “_”. Next, in step S7011, the similarity calculation unit 262 determines whether or not the dictionary notation pointer points to a symbol. Here, since the dictionary notation pointer points to the symbol, the determination is denied, and the process returns to step S7004.

  Thereafter, by repeating the third to sixth iterations, the sequence of the dictionary notation “(′ _ ｀.)” Is changed to the sequence of the main part number and the subordinate part number (S1, M1, S2, M2, M3, S3). ) (See the bottom row in FIG. 14). Note that if the arrangement of the main part number and the subordinate part number is defined instead of the emoticon dictionary notation in FIG. 11, the process of step S70 may be omitted.

  As described above, when all the processes in FIG. 13 are completed, the process proceeds to step S71 in FIG. In the subroutine of step S71, the similarity calculation unit 262 executes a process for obtaining a part number arrangement of unknown emoticons. In step S71, the similarity calculation unit 262 executes processing according to the flowchart of FIG.

  In the process of FIG. 15, in step S7101, the similarity calculation unit 262 sets an unknown notation pointer at the beginning of an unknown emoticon. Specifically, the similarity calculation unit 262 matches the unknown notation pointer with the “(” part of the emoticon string “(′ ∩ ｀.)” Input to the emoticon input receiving unit 260. Next, step S7102 is performed. Then, the similarity calculation unit 262 sets the main component pointer at the head of the main component association result, which is obtained by associating the main component number and the component as shown in FIG. Therefore, in step S7102, the main component pointer is set to “M1 ′” (see the first step in FIG. 16).

  Next, in step S7103, the similarity calculation unit 262 sets the dependent component pointer at the head of the dependent component association result. In addition, the dependent part correlation result is shown in a state ("S1 (S3)") in which the dependent part number is associated with the part as shown in FIG. Accordingly, in step S7103, the main component pointer is set to “S1 (” (see the first step in FIG. 16).

  Next, in step S7004, the similarity calculation unit 262 determines whether or not the symbols indicated by the unknown notation pointer and the main component pointer match. Here, as shown in the first level of the first time in FIG. 16, since they do not match, the determination is negative, and the process proceeds to step S <b> 7107. In step S7107, the similarity calculation unit 262 determines whether the symbols indicated by the unknown notation pointer and the subordinate component pointer match. Here, as shown in the first level of the first time in FIG. 16, since they match, the determination is affirmed and the process proceeds to step S <b> 7108.

  In step S7108, the similarity calculation unit 262 outputs the initial letter S representing the dependent component and the dependent component number (S1) to an output buffer included in the similarity calculation unit 262. In the next step S7109, the similarity calculation unit 262 advances the dependent component pointer to the next symbol. The results of steps S7108 and S7109 are shown in the first second row in FIG.

  Next, in step S7110, the similarity calculation unit 262 advances the unknown notation pointer to the next symbol. This result is shown in the first third row of FIG.

  Next, in step S7111, the similarity calculation unit 262 determines whether or not the unknown notation pointer points to a symbol. Here, since the unknown notation pointer points to the symbol, the determination is denied, and the process returns to step S7104.

  Next, the second repetition will be described. In the second repetition, as shown in the first row of the second time in FIG. 16, since the unknown notation pointer and the symbol “′” pointed to by the main part pointer match, the determination in step S7104 is affirmed, and step S7105 is performed. Migrate to In step S7105, the similarity calculation unit 262 outputs the initial value M representing the main part and the main part number (M1) included in the similarity calculation unit 262, as shown in the second row in FIG. Output to buffer.

  Next, in step S7106, the similarity calculation unit 262 advances the main component pointer to the next symbol “M2 ｀” as illustrated in the third level in the second time in FIG. Next, in step S7110, the similarity calculation unit 262 advances the unknown notation pointer to the next symbol “∩”. Next, in step S7111, the similarity calculation unit 262 determines whether or not the unknown notation pointer points to a symbol. Here, since the unknown notation pointer points to the symbol, the determination is denied, and the process returns to step S7104.

  Thereafter, by repeating the third to sixth iterations, the sequence of the input unknown emoticon expression “(′ ∩ ｀.)” Is changed to the sequence of the main part number and the subordinate part number (S1, M1, M2, M3, S3). That is, by the processing of steps S70 and S71, the similarity calculation unit 262 performs the data No. As shown in FIG. 3, the dictionary component order and the unknown emoticon association order can be obtained.

  When all the processes in FIG. 15 are completed as described above, the process proceeds to step S72 in FIG. In the subroutine of step S72, the similarity calculation unit 262 executes processing for verifying the part arrangement. In step S72, the similarity calculation unit 262 executes processing according to the flowchart of FIG. Here, as the processing order, first, the data No. Although the part arrangement is verified by using the emoticon No. 3 for the sake of convenience, the data No. described in the second row of FIG. 4 will be used to describe the processing of FIG.

  In the process of FIG. 18, first, in step S7201, the similarity calculation unit 262 sets the correspondence result pointer A at the head of the unknown face character association order. In step S7202, the similarity calculation unit 262 sets the dictionary part pointer X at the head of the dictionary part order. The state of the pointer at this stage is shown in the first row of FIG.

  Next, in step S7203, the similarity calculation unit 262 determines whether or not the correspondence result pointer A matches the part pointed to by the dictionary part pointer X. Here, as shown in the first row of FIG. 19, since they match, the determination is affirmed and the process proceeds to step S7205. If the determination in step S7203 is negative, the similarity calculation unit 262 advances the dictionary part pointer X to the next part in step S7204, and then returns to step S7203 again.

  In step S7205, the similarity calculation unit 262 sets the verification correspondence result pointer B to the part next to the current correspondence result pointer A. Next, in step S7206, the similarity calculation unit 262 determines whether or not the verification correspondence result pointer B exceeds the final part. If the determination here is affirmed, it indicates that there is no more symbol to be verified, and the processing in FIG. 18 ends. If the determination in step S7206 is negative, the process advances to step S7207. Now, as shown in the second row of FIG. 19, since the verification correspondence result pointer B has not yet exceeded the final part, the determination in step S7206 is denied and the process proceeds to step S7207. In step S7207, the verification dictionary component pointer Y is set to the component immediately after the current dictionary component pointer X. As a result of steps S7205 and S7207, the pointers are as shown in the second row of FIG. Next, in step S7208, the similarity calculation unit 262 determines whether the pointer Y has exceeded the final part in the dictionary list. Now, as shown in the second row of FIG. 19, the verification dictionary component pointer Y has not yet exceeded the final component in the dictionary list, so the determination in step S7208 is denied. If the determination here is negative, the process advances to step S7210, and the similarity calculation unit 262 determines whether or not the parts pointed to by the pointer B and Y match. Here, as shown in the second row of FIG. 19, since they match each other, the determination is affirmed and the process proceeds to step S7212. In step S7212, the similarity calculation unit 262 sets the pointer B to the next part. The state after the processing in step S7212 is shown in the third row of FIG.

  Next, in step S7213, the similarity calculation unit 262 determines whether or not the pointer B exceeds the final part. Here, since it has not exceeded, determination is denied and it returns to step S7207.

  Returning to step S7207, the similarity calculation unit 262 sets the pointer Y to the next component of X (the state of the fourth stage in FIG. 19), and after the determination in steps S7208 and S7210, in step S7211, the pointer Y Is set to the next part (the fifth stage state in FIG. 19). Then, after the determination in steps S7208 and S7210 again, the pointer Y is set to the next part in step S7211 (the sixth stage state in FIG. 19). Then, after the determinations of steps S7208 and S7210 are again made, the determination of S7210 is affirmed, and thus the process proceeds to step S7212. In step S7212, the similarity calculation unit 262 sets the pointer B to the next part (step 7 in FIG. 19).

  As described above, through the loop of steps S7207 to S7213, the part that exists on the right side of the part for which the pointer A is set in the association result is that the pointer X is set in the dictionary list. It can be confirmed whether or not it exists on the right side of the part. In this process, as shown in FIGS. 20, 21, and 22, all the combinations are confirmed while the pointers A and X are appropriately shifted in the right direction.

  Here, as shown in the eighth row of FIG. 22, when the pointers A and X are set to M3, S3 existing on the right side of the pointer A exists on the right side of the pointer X in the dictionary list. Not done. For this reason, when the pointer Y exceeds the final part of the dictionary, the process proceeds to step S7209.

  When the process proceeds to step S7209, the similarity calculation unit 262 counts the number of penalties. The penalty count represents the number of contradictory portions between the dictionary component order and the unknown emoticon association order. After step S7209 is performed as described above, the same processing as described above is repeated while setting the pointer A to the next part in step S7214, and when the pointer A exceeds the final part (step S7215 When the result is affirmed), the entire process of FIG. 18 is terminated, and the process proceeds to step S73 of FIG. In addition, data No. described in the first row of FIG. In the case of the smiley face 3, the determination in step S7208 is never affirmed, that is, the process of FIG. 18 is terminated without passing through step S7209.

When the process proceeds to step S73 in FIG. 10, the similarity calculation unit 262 performs a penalty calculation. Specifically, the similarity calculation unit 262 calculates a penalty value according to the number of penalties (that is, the number of times the process in step S7209 has been performed = the number of contradictory portions). That is, for example, when ½ is preset as the penalty value reference value and the penalty count is m, the similarity calculation unit 262 calculates the penalty value,
Penalty value = (1/2) ^m (1)
It is calculated from the formula of For example, data No. 1 in FIG. In the case of an emoticon of 3, the penalty count is 0, so the penalty value is (1/2) ⁰ = 1. In addition, data No. 1 in FIG. In the case of 4 emoticons, the penalty count is 1, so the penalty value is (1/2) ¹ = 1/2. Further, if the penalty count is 2, the penalty value is (1/2) ² = ¼.

  Next, in step S74, K is incremented by one. In step S75, the processes and determinations in steps S70 to S75 are performed until K exceeds the number of emoticons extracted in step S69B, that is, until the penalty calculation for all of the extracted emoticons is completed. repeat.

  If the determination in step S75 is affirmative, the similarity calculation unit 262 calculates the similarity in step S76A. Specifically, the similarity calculation unit 262 calculates the similarity from the product of the temporary similarity calculated in step S69A and the penalty value calculated in step S73. Here, the data No. shown in FIG. The similarity of the emoticon 3 is tentative similarity (100%) × penalty value (1) = 100%. In addition, data No. 1 shown in FIG. The similarity of the emoticon 4 is tentative similarity (60%) × penalty value (1/2) = 30%.

  In step S76B, the reading estimation unit 264 determines whether there is a degree of similarity greater than a certain threshold value (second threshold value). If the determination is negative, it is determined that the character is not an emoticon, and the character string is output to the normal text analysis unit 24, that is, the process proceeds to step S6 in FIG. On the other hand, if the determination here is affirmative, the process proceeds to step S77, and the reading estimation unit 264 acquires readings / accents of emoticons whose similarity is greater than the threshold from the emoticon dictionary of FIG. The data is output to the built-in buffer of the reading output unit 27. In this case, when there are a plurality of emoticons having a degree of similarity greater than the threshold, all the readings / accents of the plurality of emoticons are acquired and continuously output. By doing so, it is possible to read out more meaningfully. Here, it is assumed that “Gusun” is output as the emoticon reading / accent.

  In step S77, the reading estimation unit 264 may acquire readings / accents of emoticons in the emoticon dictionary having the highest similarity.

  As described above, when all the processes (step S5) in FIG. 10 are completed, the process proceeds to step S7 in FIG. In step S7, the reading output unit 267 sequentially accumulates the readings of emoticons acquired through the processing of FIG.

  Next, in step S8, the character determination unit 224 determines whether or not the input buffer 221 is empty. If the determination here is affirmed, the process proceeds to step S9, and the reading output unit 27 outputs the reading accumulated in the buffer to the voice generation unit 33.

  In step S10, the voice generation unit 33 connects the accumulated readings to generate a voice ("sad slap") and outputs it from the speaker 56 of FIG. As described above, according to the present embodiment, it is possible to estimate and read out an emoticon that is not registered in the emoticon dictionary.

  As described above in detail, according to the present embodiment, the emoticon dictionary storage unit 25 is extracted in advance as to contribute to the meaning content of the emoticon from a plurality of symbols representing the emoticon for a plurality of emoticons. The emoticon dictionary registered by associating at least one main component with the emoticon reading information is stored, and the emoticon reading estimation unit 26 (ratio acquisition unit 261) stores the emoticon dictionary in the emoticon dictionary. For each registered emoticon, the main part is compared with the part of the emoticon received by the emoticon input receiving unit 260, and the ratio of the same part as the main part in the entire inputted emoticon is calculated. get. Then, the similarity calculation unit 262 calculates the similarity between the input emoticon and the emoticon registered in the emoticon dictionary based on the ratio, and the reading estimation unit 264 calculates the similarity. Based on the emoticon readings / accents registered in the emoticon dictionary, the input emoticon readings / accents are estimated. In this way, among emoticons registered in the emoticon dictionary, the emoticon readings / accents that have a high proportion of parts that are common to the main parts are the same Thus, even if the inputted emoticon is not registered in the emoticon dictionary, the reading / accent representing the meaning can be estimated with high accuracy. This makes it possible to accurately estimate readings and accents of a wide variety of emoticons.

  In the present embodiment, the emoticon dictionary also includes subcomponents, which are symbols other than the main components, in addition to the main components, and the similarity calculation unit 262 calculates the ratio (provisional similarity) from the emoticon dictionary. ) Is greater than or equal to a predetermined threshold (first threshold), and the order of the main parts and subordinate parts of the extracted emoticons and the same parts as the main parts of the input emoticons In addition, the number of inconsistent portions (the number of penalties) between the subordinate symbols and the same component arrangement order is calculated, and the similarity is calculated based on the number of penalties and the provisional similarity. That is, of the emoticons registered in the emoticon dictionary, two conditions (the ratio of the input emoticon parts and the main parts in common is high, and the order of arrangement of the main parts and the subordinate parts is input. A reading / accent of an emoticon satisfying (similar to the arrangement order of emoticon components) is estimated as a reading / accent of the input emoticon. Therefore, even when the inputted emoticon is not registered in the emoticon dictionary, the reading / accent representing the meaning can be estimated with higher accuracy.

  In the above embodiment, the case where readings / accents are registered in the emoticon dictionary of FIG. 7 has been described, but the present invention is not limited to this. For example, only reading may be registered. Further, in the above-described embodiment, the case where emoticon reading is output from the speaker 56 by voice has been described. However, the present invention is not limited to this, and the reading of the emoticon may be output as character data, that is, for example, the emoticon may be replaced with text indicating meaning content and displayed on the display 54.

  Moreover, although the said embodiment demonstrated the case where the reading of input text (an emoticon was included) was estimated and the audio | voice was produced | generated inside the mobile telephone 100, it is not restricted to this. For example, text input to the mobile phone 100 may be transmitted to a server connected to the mobile phone 100, and reading of the input text (including emoticons) may be estimated at the server. In this case, when the server transmits the estimation result to the mobile phone 100, the mobile phone 100 can output the voice of the input text (ASP (Application Service Provider)) as in the above embodiment. In such a case, it is not necessary to provide the mobile phone 100 with a configuration for estimating the reading of emoticons, and thus the mobile phone 100 can be simplified.

  The above processing functions can be realized by a computer. In that case, a program describing the processing contents of the functions that the processing apparatus should have is provided. By executing the program on a computer, the above processing functions are realized on the computer. The program describing the processing contents can be recorded on a computer-readable recording medium.

  When the program is distributed, for example, it is sold in the form of a portable recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. It is also possible to store the program in a storage device of a server computer and transfer the program from the server computer to another computer via a network.

  The computer that executes the program stores, for example, the program recorded on the portable recording medium or the program transferred from the server computer in its own storage device. Then, the computer reads the program from its own storage device and executes processing according to the program. The computer can also read the program directly from the portable recording medium and execute processing according to the program. Further, each time the program is transferred from the server computer, the computer can sequentially execute processing according to the received program.

  The above-described embodiment is an example of a preferred embodiment of the present invention. However, the present invention is not limited to this, and various modifications can be made without departing from the scope of the present invention.

2 Language analysis unit (emoticon reading information estimation device)
3 Waveform processing unit (output device)
25 Emoticon dictionary storage (storage)
260 emoticon input reception part (reception part)
261 Ratio acquisition unit (acquisition unit)
262 similarity calculation unit (calculation unit)
H.264 Reading Estimator (Estimator)

Claims (8)

An emoticon reading information estimation device that estimates emoticon reading information represented by a plurality of symbols,
Regarding a plurality of emoticons, at least one main symbol extracted in advance as contributing to the meaning content of the emoticon from a plurality of symbols representing the emoticon and the reading information of the emoticon are registered in association with each other A storage unit for storing the emoticon dictionary;
A reception unit that accepts input of emoticons from outside,
For each emoticon registered in the emoticon dictionary, the main symbol is compared with a plurality of symbols representing the emoticon input to the receiving unit, and a plurality of all symbols representing the inputted emoticon An acquisition unit for acquiring a proportion of the same symbol as the main symbol in
Based on the ratio acquired by the acquisition unit, a calculation unit that calculates the similarity between the input emoticon and the emoticon registered in the emoticon dictionary;
An emoticon reading information estimation device comprising: an estimation unit that estimates the input emoticon reading information from emoticon reading information registered in the emoticon dictionary based on the similarity. In the emoticon dictionary stored in the storage unit, the reading of the emoticon and the accent are registered as the reading information of the emoticon,
The estimation unit estimates the reading and accent of the inputted emoticon from the emoticon reading and accent registered in the emoticon dictionary based on the similarity. The emoticon reading information estimation apparatus according to Item 1. In the emoticon dictionary stored in the storage unit, along with the at least one main symbol, a subordinate symbol that is a symbol other than the main symbol among a plurality of symbols representing the emoticon is also registered,
The calculation unit extracts the emoticon whose ratio is equal to or greater than a predetermined first threshold from the emoticon dictionary, the order of arrangement of the main symbols and subordinate symbols of the extracted emoticon, and the input Calculate the number of inconsistent portions with the same symbol as the main symbol of the emoticon and the arrangement order of the same symbols as the subordinate symbols, and calculate the similarity based on the number of the inconsistent portions and the ratio The emoticon reading information estimation apparatus according to claim 1 or 2, wherein   The said estimation part estimates the reading information of the emoticon with the largest similarity calculated by the said calculation part as reading information of the said inputted emoticon, The any one of Claims 1-3 characterized by the above-mentioned. The emoticon reading information estimation apparatus according to the item.   The estimation unit estimates all the emoticon reading information whose similarity calculated by the calculation unit is equal to or greater than a predetermined second threshold as the input emoticon reading information. The emoticon reading information estimation apparatus as described in any one of Claims 1-3. The emoticon reading information estimation device according to any one of claims 1 to 5,
An information terminal comprising: an output device that outputs the input reading information of the emoticon estimated by the estimation unit of the emoticon reading information estimation device as character data or speech. An emoticon reading information estimation method for estimating emoticon reading information represented by a plurality of symbols using a computer,
A reception process for receiving input of emoticons from the outside;
Regarding a plurality of emoticons, at least one main symbol extracted in advance as contributing to the meaning content of the emoticon from a plurality of symbols representing the emoticon and the reading information of the emoticon are registered in association with each other With reference to the emoticon dictionary, for each emoticon registered in the emoticon dictionary, the main symbol is compared with a plurality of symbols representing the inputted emoticon, and the inputted emoticon is An acquisition step of acquiring a ratio of the same symbol as the main symbol in all of a plurality of symbols to be represented;
A calculation step of calculating a similarity between the input emoticon and the emoticon registered in the emoticon dictionary based on the ratio acquired in the acquisition step;
The computer executes an estimation step of estimating the input emoticon reading information from the emoticon reading information registered in the emoticon dictionary based on the similarity. To estimate emoticon reading information. An emoticon reading information estimation program for causing a computer to estimate reading information of emoticons represented by a plurality of symbols,
On the computer,
Acceptance process to accept external emoticon input,
Regarding a plurality of emoticons, at least one main symbol extracted in advance as contributing to the meaning content of the emoticon from a plurality of symbols representing the emoticon and the reading information of the emoticon are registered in association with each other With reference to the emoticon dictionary, for each emoticon registered in the emoticon dictionary, the main symbol is compared with a plurality of symbols representing the inputted emoticon, and the inputted emoticon is An acquisition step of acquiring a ratio of the same symbol as the main symbol in all of a plurality of symbols to be represented;
Based on the ratio acquired in the acquisition step, a calculation step of calculating a similarity between the inputted emoticon and the emoticon registered in the emoticon dictionary, and based on the similarity, An emoticon reading information estimation program for executing an estimation step of estimating the input emoticon reading information from emoticon reading information registered in an emoticon dictionary.

Images