JP2009169513A - Device, method and program for estimating nickname - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a nickname estimation device for more properly acquiring a nickname from such designation as a person's name. <P>SOLUTION: The nickname estimation device includes: a rule storage part 121 for storing the generation rule of the candidates of a nickname including position information indicating the positions of characters included in the candidates of the nickname among characters included in designation and a predetermined additional character string; a designation input part 101 for inputting designation; a candidate generation part 102 for acquiring the characters of the positions shown by the position information of the generation rule among the characters included in the input designation, and for generating the candidates of the nickname by connecting the acquired characters and the additional character string of the generation rule; and an output part 103 for outputting the generated candidates of the nickname. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an apparatus, a method, and a program for estimating and outputting a nickname using a name as an input.

  In recent years, information retrieval technology has been widely used and used in various situations. Generally, when performing an information search, a search keyword is given as an input, and a document that matches or is strongly related to the input search keyword is returned as a result. For example, using a person's name as a search keyword, a document related to the person is searched from a Web page.

  In addition to the official name, names such as personal names include nicknames, abbreviations, aliases, and common names as other names. For this reason, if the correspondence between the official name and the nickname is not known, only one is searched as a search keyword, and it is difficult to search for necessary information comprehensively. For example, when there is a person with the personal name “Oda Ymi” and the nickname “YRin”, if only “Oda Ymi” is the search keyword, information including only “YRin” cannot be acquired.

  On the other hand, human interface using speech recognition technology is also being put into practical use. In the future, it is assumed that information retrieval by voice dialogue or voice will be performed. In the information search process by voice, there is a high possibility that expressions such as nicknames are used compared to the information search process by text. For this reason, in order to promote a user's free speech, it is necessary to be able to recognize expressions such as nicknames in association with formal names.

  Generally, in speech recognition, a speech recognition dictionary for collating with spoken speech is used. An unregistered word that does not exist in the speech recognition dictionary cannot be recognized by speech. For this reason, if a nickname or the like is not registered in the speech recognition dictionary, the nickname spoken by speech cannot be recognized. Furthermore, since the names and nicknames of persons who can be searched can be updated every day, it is necessary to update the speech recognition dictionary so that the names and the like that can be updated can be recognized normally, but the cost is high.

  For example, a person to be searched includes an entertainer who appears in a television program. However, new entertainers appear in television programs and the like, and their nicknames are often spread rapidly. In order to cope with names and the like that can be updated every day in this way, it is necessary to continue to update the speech recognition dictionary.

  In addition, as a technique for acquiring the correspondence between abbreviations and official names, the input words are divided into words, the combination of the initials of the divided words, and the official names and abbreviations of public institutions, large corporations, etc. Has been proposed (for example, Patent Document 1). However, as in the case of the above-described speech recognition dictionary, when a person name that can be updated daily is targeted, the cost for continuing to provide the latest database manually increases. In addition, a simple abbreviation generation method that only combines the initial letters of words cannot generate nicknames that can include characters other than words included in official names. For example, although the abbreviation “N Benren” can be generated from the official name “N Lawyer Federation”, a nickname that can include characters other than the words included in the official name “N Lawyer Federation” cannot be generated.

  Patent Document 2 proposes a technique for generating an abbreviation from a name using an abbreviation generation rule. In this method, an abbreviation generation rule that specifies a location used for abbreviation generation among the basic words obtained by dividing the input word is held. Then, abbreviation candidates are generated from the input words by this abbreviation generation rule. Further, if there is an abbreviation candidate in a keyword assigned to the target document of the word search device, it is determined as an abbreviation, and a word search using the abbreviation is also performed.

  The abbreviation is an abbreviation for a long name, and is composed of only a character string related to the official name. For example, in the case of a person having a personal name “P field Aya”, a personal name reading “Pita Eiya”, and a nickname “Pii A”, it can be said that an abbreviation composed only of a character string included in the personal name reading is a nickname. On the other hand, an expression completely different from the personal name may be nicknamed, such as a person with the personal name “Suda Uro” and the nickname “Tokei Prince”. In addition, for a person who has a person name “Ida Miya” and a person name reading “Aitametsuya”, a character that is not related to the original person name is inserted using a part of the person name reading, and a new word is used. A nickname “Metchy” may be created. In the latter two cases, the nickname cannot be generated by the abbreviation generation method as in Patent Document 2.

  In Patent Document 2, a keyword is assigned to a target document of a word search device, and if an abbreviation candidate exists in a word given as a keyword, the abbreviation candidate is selected as an abbreviation. For example, if there are abbreviation candidates such as “life insurance”, “life insurance”, and “life insurance” for the search keyword “life insurance”, if “life insurance” exists as a keyword assigned to the document, only “life insurance” Certified as an abbreviation.

  That is, in Patent Document 2, it is assumed that there is a document to which abbreviation candidates are assigned as keywords. However, for example, when there is no data in which the nickname is divided into keywords as a keyword as in the case of searching from a Web page, an appropriate nickname candidate is selected from the generated nickname candidate after the nickname candidate is generated. It becomes difficult to select.

  On the other hand, Non-Patent Document 1 proposes a technique for extracting a name by using the fact that the name and the formal name are represented by a typical Japanese expression “(name) thing (formal name)”. Yes. Specifically, in the method of Non-Patent Document 1, a web search is performed using “ko (formal name)” as a search keyword, and a name corresponding to the formal name is extracted from characters appearing before “ko”. . According to this method, there is a possibility that any nicknames of the above-mentioned “Peiei”, “Prince Tokei”, and “Mechi” may be acquired.

JP 2003-333161 A Japanese Patent Laid-Open No. 11-25117 Tomoko Tooma et al., “Extracting Person Names Using Web Data”, DBSJ Letters Vol.5 No.2

  However, even with the method of Non-Patent Document 1, a correct nickname corresponding to the official name may not be acquired. In Non-Patent Document 1, a morphological analyzer is used to extract a nominal part from a character string that appears before “Koto”. For this reason, when the nickname is not registered as a word in the dictionary used by the morphological analyzer and the nickname is a coined word made from the name, it becomes difficult to separate the nickname from the character string. . For example, in the method of Non-Patent Document 1, even if the correct nickname is “U-chan”, “-chan” may be extracted as the nickname.

  Further, in Non-Patent Document 1, a typical expression such as “(nickname) thing (official name)” is used as described above, but the nickname is not described in such a typical expression in the search target document. In case you can't get a nickname.

  The present invention has been made in view of the above, and an object thereof is to provide an apparatus, a method, and a program that can more appropriately acquire a nickname from a name such as a person's name.

  In order to solve the above-described problems and achieve the object, the present invention is a nickname estimation device that estimates a nickname of a name from a name, and includes characters included in the nickname candidate among characters included in the name Included in the input name, a rule storage unit that stores generation information of the nickname candidate including position information representing the position of the character and a predetermined additional character string, a name input unit that inputs the name A generation unit that acquires a character at a position represented by the position information of the generation rule, and generates the nickname candidate that combines the acquired character and the additional character string of the generation rule; And an output unit that outputs the generated nickname candidates.

  Further, the present invention is a method and program capable of executing the above-described apparatus.

  According to the present invention, there is an effect that a nickname can be acquired more appropriately from a name such as a person's name.

  Exemplary embodiments of an apparatus, a method, and a program according to the present invention will be described below in detail with reference to the accompanying drawings.

(First embodiment)
As described above, conventionally, nicknames, abbreviations, aliases, and the like have been manually created and registered in a database. In the case of an abbreviation, a method of estimating an abbreviation from a formal name using a predetermined abbreviation generation pattern has been used. However, information that is updated daily, such as program guide data (EPG (Electronic Program Guide) data) of television programs, needs to be updated frequently due to the appearance of new entertainers, for example. For this reason, it is difficult to respond sufficiently only by storing it in the dictionary. In addition, it is expensive to manually update the dictionary.

  The nickname estimation apparatus according to the first embodiment generates a nickname candidate for an input name in accordance with a predetermined nickname candidate (nickname candidate) generation rule (nickname generation rule).

  FIG. 1 is a block diagram illustrating a configuration of a nickname estimation apparatus 100 according to the first embodiment. As shown in FIG. 1, the nickname estimation device 100 includes a rule storage unit 121, a name input unit 101, a candidate generation unit 102, and an output unit 103.

  The rule storage unit 121 stores a nickname generation rule for generating a nickname candidate for the input name. FIG. 2 is a diagram illustrating an example of a nickname generation rule stored in the rule storage unit 121. As shown in FIG. 2, the nickname generation rule includes information (symbols) for specifying a nickname to be generated using a character at which position of which character type among the input names, and a part of the nickname. It includes additional character strings (such as “chan”) that make up.

  In the present embodiment, it is possible to specify the character type and the character position by using a three-digit numerical symbol. The numerical value of the hundreds represents the character type, with “1” corresponding to the official name, “2” corresponding to hiragana notation, and “3” corresponding to katakana notation. The numerical value at the 10's position represents information for identifying the position of each word unit constituting the name with respect to the beginning of the name. For example, in the case of a person name whose first and last names are word units, if the numerical value at the 10th place is “1”, it represents the last name, and “2” represents the first name. For example, in a name such as a group name or an organization name, there are cases where the name is composed of three or more word units, so that the numerical value at the tenth place can be three or more. In the following, a case where the numerical value at the 10th place takes either “1 (last name)” or “2 (first name)” will be described as an example. The numerical value of 1's represents information for identifying the position from the beginning in each word unit.

  In the figure, an example in which the nickname generation rule is applied is described to the right of the nickname generation rule column, but only the nickname generation rule is stored in the actual rule storage unit 121.

  Further, the rule storage unit 121 can be configured by any generally used storage medium such as an HDD (Hard Disk Drive), an optical disk, a memory card, and a RAM (Random Access Memory).

  The name input unit 101 receives an input of a name such as a person name. Note that the name input unit 101 receives an input of a name in a format that can identify a word unit. For example, in the case of a person name, the name input unit 101 inputs a family name and a first name independently.

  In addition, the name input unit 101 accepts input of hiragana notation and katakana notation of the official name together with the official name. For example, when the personal name is “P field Aya”, the name input unit 101, together with the last name “P field” and the first name “A field”, which is an official name in which the first name and the last name are separated, the hiragana notation “Pita”. It accepts the input of “Eiya” as the first name and “Pita” as the last name in katakana and “Eiya” as the first name.

  Note that the method for inputting a name is not limited to this, and any method can be applied as long as it is a method that can identify a word unit constituting a name. For example, the name input unit 101 may be configured to accept an input of a name divided in units of words by a predetermined character string such as a space.

  Moreover, you may comprise so that a name may be input, without dividing | segmenting into a phrase unit. In this case, for example, the name input unit 101 may be configured to divide “P field Aya” input using a personal name dictionary into a family name “P field” and a name “A field”. In addition, instead of accepting input of hiragana notation and katakana notation, the name input unit 101 is configured to estimate the reading of the official name input using a personal dictionary and acquire hiragana notation and katakana notation. May be.

  The candidate generation unit 102 refers to the nickname generation rule stored in the rule storage unit 121 and generates a nickname candidate for the input name. Specifically, the candidate generation unit 102 converts the input name into a symbol, and replaces the same symbol as the converted symbol in the nickname generation rule with a character corresponding to the converted symbol. Generate nickname candidates.

  Here, the specific example of the conversion method from a name to a symbol is demonstrated using FIGS. FIG. 3 is a diagram illustrating an example of input name data. As shown in FIG. 3, hereinafter, a case where a name 1 that is an official name including kanji, a name 2 that is hiragana notation, and a name 3 that is katakana notation will be described as an example. In addition, the symbol column of the figure represents the result of converting each name into a symbol. Details of the conversion method will be described with reference to FIGS.

  FIG. 4 is a diagram illustrating an example of converting the name 1 (official name) in FIG. 3 into a symbol. FIG. 4 shows, for each character notation included in the name 1, the character type, the position in the phrase unit (last name or first name), the character position in the phrase unit, and the symbol that is the conversion result. Since name 1 is an official name, the character type is “1”. Also, the first and last name field corresponding to each character notation is set to either “1” or “2” depending on whether it is located before or after the space as a delimiter. Furthermore, the character position of each character notation in the last name or the first name is set. Then, a three-digit numerical value obtained by combining the numerical values corresponding to the character type, first name, and character position is generated as a symbol of the conversion result.

  FIG. 5 is a diagram illustrating an example in which name 2 (hiragana notation) in FIG. 3 is converted into a symbol. Since name 2 is written in hiragana, “2” is set as the character type. FIG. 6 is a diagram illustrating an example of converting the name 3 (katakana notation) in FIG. 3 into a symbol. Since name 3 is in katakana notation, “3” is set as the character type.

  The candidate generation unit 102 thus converts each name into a symbol, and generates a symbol as shown in FIG. Then, the candidate generation unit 102 generates a nickname candidate by applying a character corresponding to the converted symbol to a nickname generation rule including the same symbol as the converted symbol.

  For example, the person name 301 (“P field Aya”) and the katakana notation 302 (“Pita Aiya”) in FIG. 3 are input as formal names, and the nickname generation rule “311 312 321 322” at the top of FIG. On the other hand, consider a case where the input name is applied. In this case, since all the symbols in the nickname generation rule have the number of the 100's place, the candidate generation unit 102 generates a nickname candidate by applying the katakana character. Specifically, the candidate generation unit 102 includes “pi” in the “311” portion, “i” in the “312” portion, and “e” in the “321” portion in the nickname generation rule at the top of FIG. ”And“ 322 ”are applied to“ i ”to generate a nickname candidate 201 (“ PIA ”) in FIG. 2.

  The output unit 103 outputs one or more nickname candidates generated by the candidate generation unit 102. If no matching nickname generation rule exists and no nickname candidate is generated, the output unit 103 may output a result indicating that there is no corresponding candidate.

  Next, the nickname estimation process by the nickname estimation apparatus 100 according to the first embodiment configured as described above will be described with reference to FIG. FIG. 7 is a flowchart showing the overall flow of the nickname estimation process in the first embodiment.

  First, the name input unit 101 inputs name data including an official name, hiragana notation, and katakana notation (step S701). Next, the candidate generation unit 102 symbolizes the formal name, hiragana notation, and katakana notation in the name data (step S702).

  Next, the candidate generation unit 102 acquires a nickname generation rule from the rule storage unit 121 (step S703). Then, the candidate generation unit 102 generates a nickname candidate by applying the symbolized name to the nickname generation rule (step S704). Specifically, for each acquired nickname generation rule, candidate generation unit 102 converts the same symbol as the symbol converted from the name in step S702 among the symbols included in the nickname generation rule, as the conversion source of the converted symbol. Generate nickname candidates replaced with letters. Finally, the output unit 103 outputs the generated nickname candidate (step S705) and ends the nickname estimation process.

  As described above, in the nickname estimation device according to the first embodiment, a nickname candidate for an input name can be generated according to a predetermined nickname generation rule, and thus a nickname is more appropriately acquired from a name such as a person name. be able to.

  Moreover, since the speech recognition dictionary and the nickname dictionary can be updated with the generated nickname candidates, it is possible to reduce manual dictionary creation costs. Furthermore, in the information search, by using the correspondence between the official name and the generated nickname candidate, even if either one is input, it is possible to perform a search using both the nickname and the official name. It becomes possible.

  On the other hand, if a nickname is added to the speech recognition dictionary based on the nickname estimation result during voice conversation, the nickname can be recognized correctly even if the user names the nickname. become. Further, even when a person's name is nicknamed, it is possible to correctly understand the name of the name by converting the nickname to an official name.

(Second Embodiment)
The nickname estimation device according to the second embodiment searches for a generated nickname candidate for a document acquired from the Web or the like, and selects a nickname candidate included in the document, thereby further appropriately nickname. Get candidates.

  FIG. 8 is a block diagram showing a configuration of a nickname estimation apparatus 800 according to the second embodiment. As illustrated in FIG. 8, the nickname estimation apparatus 800 includes a rule storage unit 121, a document storage unit 822, a name input unit 101, a candidate generation unit 102, an output unit 103, and a selection unit 804. Yes.

  The second embodiment is different from the first embodiment in that a document storage unit 822 and a selection unit 804 are added. Other configurations and functions are the same as those in FIG. 1 which is a block diagram showing the configuration of the nickname estimation apparatus 100 according to the first embodiment, and thus the same reference numerals are given and description thereof is omitted here.

  The document storage unit 822 stores a document that is a processing result of a search process that is executed using a predetermined document group such as a Web page as a search target and an official name as a search keyword. The document storage unit 822 stores a predetermined number of higher-order documents among the search result documents ranked according to the number of search keywords, for example.

  The selection unit 804 selects a more appropriate nickname candidate from the nickname candidates generated by the candidate generation unit 102. Specifically, the selection unit 804 first determines, for each generated nickname candidate, whether or not a nickname candidate is included in the document stored in the document storage unit 822. Then, the selection unit 804 executes a search for a nickname candidate included in the document in the document storage unit 822 using a collocation “(nickname candidate) thing (formal name)” as a search keyword and a Web page as a search target. To do. Then, the selection unit 804 ranks nickname candidates in descending order of the searched Web pages, and selects a predetermined number of nickname candidates at the top.

  Next, a nickname estimation process performed by the nickname estimation apparatus 800 according to the second embodiment configured as described above will be described with reference to FIG. FIG. 9 is a flowchart showing the overall flow of the nickname estimation process in the second embodiment.

  The name input process, symbolization process, rule acquisition process, and candidate generation process from step S901 to step S904 are the same processes as from step S701 to step S704 in the nickname estimation apparatus 100 according to the first embodiment. The description is omitted.

  After the nickname candidate is generated, the selection unit 804 searches a predetermined document group such as a Web page using the official name as a search keyword, and stores a document as a predetermined number of higher-order search results in the document storage unit 822 ( Step S905).

  Next, the selection unit 804 determines whether or not each nickname candidate exists in the saved document, and selects only the nickname candidate included in the saved document (step S906). Further, the selection unit 804 performs a search on the Web for each of the selected nickname candidates using the collocation “(nickname candidate) thing (formal name)” as a search keyword (step S907). Then, the selection unit 804 arranges nickname candidates in the order of the number of hits, and selects the top N candidates (step S908).

  Finally, the output unit 103 outputs the selected nickname candidate (step S909) and ends the nickname estimation process.

  Note that the selection unit 804 may be configured to select a nickname candidate designated by the user from the nickname candidates output by the output unit 103. In this case, for example, the output unit 103 displays the generated nickname candidate on a display device such as a display, and allows the user to specify a desired nickname candidate from the displayed nickname candidates using an interface such as a keyboard or a mouse. . Further, for example, the output unit 103 is configured to synthesize the generated nickname candidate text data with a voice signal and output it from a speaker, and the user can specify a desired nickname candidate by voice through an interface such as a microphone. Also good.

  When this embodiment is applied to an interactive device such as a robot that calls a user, a name of the user or the like may be input by the name input unit 101 and the nickname output from the output unit 103 may be used as a name expression. . In this case, the interactive apparatus may be configured to arbitrarily select a nickname candidate by using the selection unit 804 and to confirm with a user whether or not the nickname may be called with the selected nickname.

  As described above, in the nickname estimation apparatus according to the second embodiment, a generated nickname candidate is searched for a document acquired from a predetermined document group such as the Web, and the nickname candidate included in the document is determined. By selecting, a more appropriate nickname candidate can be acquired.

(Third embodiment)
The nickname estimation apparatus according to the third embodiment learns a nickname generation rule by inputting a pair of a name and a nickname known in advance for the name.

  FIG. 10 is a block diagram illustrating a configuration of a nickname estimation apparatus 1000 according to the third embodiment. As illustrated in FIG. 10, the nickname estimation apparatus 1000 includes a rule storage unit 121, a name input unit 101, a candidate generation unit 102, an output unit 103, a learning data input unit 1005, and a learning unit 1006. ing.

  The third embodiment is different from the first embodiment in that a learning data input unit 1005 and a learning unit 1006 are added. Other configurations and functions are the same as those in FIG. 1 which is a block diagram showing the configuration of the nickname estimation apparatus 100 according to the first embodiment, and thus the same reference numerals are given and description thereof is omitted here.

  The learning data input unit 1005 inputs learning data in which a name is associated with a nickname already known for the name. The learning data input unit 1005 inputs pairs of different notations such as formal names divided into first and last names, hiragana representations of official names, and katakana representations of official names, and nicknames as learning data. Note that a pair of a name before division and its nickname may be given to the first and last names.

  The learning unit 1006 generates a new nickname generation rule from the input learning data, and stores the generated nickname generation rule in the rule storage unit 121. Specifically, the learning unit 1006 first acquires a common character that is included in both the name and nickname of the input learning data. Then, the learning unit 1006 symbolizes the obtained common character by the same method as the candidate generation unit 102. Further, the learning unit 1006 creates a nickname generation rule by replacing common characters in nicknames of learning data with symbols of name data as shown in FIG.

  Next, learning processing by the nickname estimation apparatus 1000 according to the third embodiment configured as described above will be described with reference to FIGS. 11 and 12. FIG. 11 is a flowchart illustrating the entire flow of the learning process according to the third embodiment. FIG. 12 is a diagram illustrating an example of the generated nickname generation rule.

  Note that the nickname estimation process based on the learned nickname generation rule or the nickname generation rule stored in advance is the same as that in FIG. 7 representing the nickname estimation process of the first embodiment, and a description thereof will be omitted.

  First, the learning data input unit 1005 inputs learning data that is a pair of a name and a nickname (step S1101). Next, the learning unit 1006 acquires a common character common to the name and the nickname (step S1102). Next, the learning unit 1006 symbolizes the acquired common character (step S1103).

  For example, when learning data in which the official name 1201 (“Imi Hada”) in FIG. 12, which is a name divided into first and last names, is associated with the nickname 1202 (“I-chan”) is input, the learning unit 1006 Obtains “I” as a common character. Since “I” is the first character of the name of the official name, the learning unit 1006 symbolizes this common character as “121”.

  Next, the learning unit 1006 creates a nickname generation rule by combining the symbol of the common character and a character other than the common character in the nickname (step S1104). In the above example, the nickname generation rule 1203 (“121-chan”) is created by combining the symbolized “121” and characters other than the common character “I” in the nickname.

  In FIG. 12, nickname generation generated when learning data that is a pair of the hiragana notation 1211 (“Pii no Esya”) of the person name “P No Saya” and its nickname 1212 (“Piichan”) is input. An example of rule 1213 (“211 212-chan”) is also shown.

  Note that the learning unit 1006 further generates a new nickname generation rule by comparing the nickname generation rules with each other. Specifically, the learning unit 1006 first searches the rule storage unit 121 for another nickname generation rule (referred to as rule 2) in which only a symbol part is replaced with a certain nickname generation rule (referred to as rule 1). To do. When such a rule 2 exists, the learning unit 1006 further searches for another nickname generation rule (referred to as rule 3) having the same symbol part as that of rule 1, and determines the symbol part of rule 3 as rule 2. A new nickname generation rule replaced with the symbol part is generated.

  For example, as shown in FIG. 12, there are three nickname generation rules: a nickname generation rule 1203 (“121-chan”), a nickname generation rule 1213 (“211 212-chan”), and a nickname generation rule 1223 (“121 Yan”). Suppose you can learn directly from the learning data. Among these, the nickname generation rule 1203 (“121-chan”) and the nickname generation rule 1213 (“211 212-chan”) differ only in the portions “121” and “211 212” that are symbols. For the nickname generation rule 1203 (“121-chan”), there is another nickname generation rule 1223 (“121 Yan”) that includes the same symbol “121”. Therefore, the learning unit 1006 newly generates a nickname generation rule (“211 212 yan”) in which the part of the symbol “121” of the nickname generation rule 1223 (“121 yan”) is replaced with “211 212”. Can do.

  As described above, even if there is no data that can be directly learned in the learning data, another nickname generation rule (“211 212 Yan”) can be analogically learned from the nickname generation rule that has already been learned. it can.

  Returning to FIG. 11, the learning unit 1006 stores the generated nickname generation rule in the rule storage unit 121 (step S1105), and ends the learning process.

  FIG. 13 is a diagram illustrating a usage example of the learned nickname generation rule. FIG. 13 shows an example in which an official name 1301 (“P field Y child”), a hiragana notation 1302 (“Pita Waiko”) and a katakana notation 1303 (“Pita Waiko”) are input. Is shown. In this case, nickname candidates 1311, 1312, and 1313 ("Y-chan" and "Pii-chan") are obtained from the three nickname generation rules ("121-chan", "211 212-chan", and "121-yan") that are directly learned. , “Y Yan”) is generated. Furthermore, a nickname candidate 1314 (“Piiyan”) can be further generated from the nickname generation rule (“211 212 Yan”) learned by analogy from the directly learned nickname generation rule.

  Thus, in the nickname estimation apparatus according to the third embodiment, a nickname generation rule can be learned using learning data in which a name and a nickname are associated with each other.

  As described above, since the abbreviation is composed only of the character string related to the original name, it is relatively easy to write down the abbreviation generation rule for generating the abbreviation as in Patent Document 2. On the other hand, the nickname has a variety of expressions because characters unrelated to the name are inserted. For this reason, it may be difficult to write down the rules. If the nickname generation rule can be learned by the method of the present embodiment, such a problem can be solved.

(Fourth embodiment)
The nickname estimation apparatus according to the fourth embodiment extracts a nickname candidate from external data such as the Web in addition to generation of a nickname candidate based on a nickname generation rule, and selects an appropriate nickname candidate from the generated and extracted nickname candidates To do.

  FIG. 14 is a block diagram showing a configuration of a nickname estimation apparatus 1400 according to the fourth embodiment. As illustrated in FIG. 14, the nickname estimation device 1400 includes a rule storage unit 121, a document storage unit 822, a name input unit 101, a candidate generation unit 102, an output unit 103, a selection unit 1404, and a candidate extraction unit. 1407.

  In the fourth embodiment, the candidate extraction unit 1407 is added and the function of the selection unit 1404 is different from that of the second embodiment. Other configurations and functions are the same as those in FIG. 8, which is a block diagram showing the configuration of the nickname estimation apparatus 800 according to the second embodiment, and thus are denoted by the same reference numerals and description thereof is omitted here.

  The candidate extraction unit 1407 extracts a character string that becomes a nickname candidate from external data such as data on the Web. The candidate extraction unit 1407 uses a typical expression such as “(nickname) thing (official name)” to search for a character string including a nickname candidate from external data. Specifically, the candidate extraction unit 1407 searches external data for a certain name using “ko (formal name)” as a search keyword. Then, the candidate extraction unit 1407 acquires a character string including a predetermined number of characters before “Koto (official name)” from the obtained document, and extracts a nickname candidate from the acquired character string. Details of the method for acquiring the character string and the method for extracting the nickname candidate from the acquired character string will be described later.

  The selection unit 1404 performs nickname candidate selection processing on the nickname candidates extracted by the candidate extraction unit 1407 in addition to the nickname candidates generated by the candidate generation unit 102. Further, the selection unit 1404 determines whether or not a nickname candidate is included in a predetermined number of characters before and after the official name in the document stored in the document storage unit 822 for each generated nickname candidate. Then, the selection unit 1404 searches the Web page for a nickname candidate included within a predetermined number of characters before and after the official name in the document using the collocation “(nickname candidate) (formal name)” as a search keyword. Perform the search as Then, the selection unit 1404 ranks nickname candidates in descending order of the searched Web pages, and selects a predetermined number of nickname candidates at the top.

  Next, a nickname estimation process performed by the nickname estimation apparatus 1400 according to the fourth embodiment configured as described above will be described with reference to FIG. FIG. 15 is a flowchart showing the overall flow of the nickname estimation process in the fourth embodiment.

  The name input process, symbolization process, rule acquisition process, and candidate generation process from step S1501 to step S1504 are the same processes as from step S901 to step S904 in the nickname estimation apparatus 800 according to the second embodiment. The description is omitted.

  After the nickname candidate is generated, the selection unit 1404 searches a predetermined document group such as a Web page using the official name as a search keyword, and a document (hereinafter referred to as a document page) that is a predetermined upper number of search results is a document. It preserve | saves at the memory | storage part 822 (step S1505).

  Next, the selection unit 1404 obtains a character string of s characters before and after the official name (s is an integer of 1 or more) from the saved document page (step S1506). Furthermore, the selection unit 1404 selects only nickname candidates that appear in the acquired character string (step S1507).

  Next, the candidate extraction unit 1407 executes nickname candidate extraction processing for extracting nickname candidates from the document page (step S1508). Details of the nickname candidate extraction process will be described later.

  Next, the selection unit 1404 searches the Web for the nickname candidate selected in step S1507 and the nickname candidate extracted in step S1508 using the collocation “(nickname candidate) thing (official name)” as a search keyword. Is performed (step S1509). Then, the selection unit 1404 arranges nickname candidates in the order of the number of hits, and selects the top N candidates (step S1510).

  Finally, the output unit 103 outputs the selected nickname candidate (step S1511) and ends the nickname estimation process.

  Next, details of the nickname candidate extraction process in step S1508 will be described with reference to FIG. FIG. 16 is a flowchart showing the overall flow of the nickname candidate extraction process in the fourth embodiment.

  First, the candidate extraction unit 1407 performs a Web search using “ko (formal name)” as a search keyword, and acquires the top N documents (step S1601). Next, the candidate extraction unit 1407 acquires, from the acquired document, a character string str of t characters (t is an integer of 1 or more) before the search keyword “ko (formal name)” (step S1602). .

  Next, the candidate extraction unit 1407 generates, for each of the acquired character strings str, a suffix that is a character string in a range from a character at an arbitrary position of the character string str to the last character of the character string str (Step S1). S1603).

  FIG. 17 is a diagram illustrating an example of a suffix. FIG. 17 shows eight suffixes 1711 to 1718 created from the character string 1701 (“Today is a good weather”) (“Da”, “Dai”, “Weather”, “I have a good weather”, “Good” Weather "," is good weather "," day is good weather "," today is good weather ").

  Returning to FIG. 16, the candidate extraction unit 1407 acquires, for each suffix, one character before the suffix from the stored document page and the character string str, and acquires the number of types of the acquired characters (Step). S1604).

  FIG. 18 is a schematic diagram illustrating an example of a method for acquiring the number of types. FIG. 18 shows only the three character strings “Today is good weather”, “Tomorrow is good weather”, and “Pleasant weather” for the suffix of FIG. It represents the number of types of characters before each suffix when present in str. In addition, the numerical value in FIG. 18 represents the number of types of characters immediately before each suffix. Further, when counting the number of types, the katakana notation in the document page, the character string str, and the suffix is replaced with the hiragana notation.

  In this example, for the suffix 1715 in FIG. 17 (“good weather”), for example, a character 1801 (“ha”) and a character 1802 (“no”) are acquired as the immediately preceding characters. Therefore, the number of types for the suffix 1715 is two.

  Returning to FIG. 16, the candidate extraction unit 1407 acquires an unprocessed suffix from the suffixes (step S1605). Then, the candidate extraction unit 1407 acquires the suffix obtained in step S1604 for the suffix obtained in step S1604 and the suffix obtained by deleting the first character of the suffix in step S1604. It is determined whether the number of types is 1 (step S1606).

  If the number of types acquired for the suffix is greater than 1 and the number of types acquired for the suffix from which the first character of the suffix is deleted is 1 (step S1606: YES), candidate extraction The unit 1407 extracts the acquired suffix as a nickname candidate (step S1607).

  If the number of types acquired for the suffix is greater than 1 and the number of types acquired for the suffix from which the first character of the suffix is deleted is not 1 (step S1606: NO), the candidate extraction unit 1407 further determines whether the number of types acquired for the suffix is 1 and the number of types acquired for the suffix from which the first character of the suffix is deleted is greater than 1 (Step S1608).

  If the number of types acquired for the suffix is 1 and the number of types acquired for the suffix from which the first character of the suffix is deleted is greater than 1 (step S1608: YES), candidate extraction The unit 1407 extracts a suffix obtained by deleting the first character from the acquired suffix as a nickname candidate (step S1609).

  Next, the candidate extraction unit 1407 determines whether all suffixes have been processed (step S1610). If all suffixes have not been processed (step S1610: NO), the candidate extraction unit 1407 acquires the next unprocessed suffix and repeats the processing (step S1605).

  When all the suffixes have been processed (step S1610: YES), the candidate extraction unit 1407 deletes nickname candidates that are partial character strings of other nickname candidates with the same frequency in the character string str (step S1611). .

  By such processing, an appropriate nickname candidate can be extracted from a document searched for “ko (formal name)” as a search keyword. For example, as compared to the case of using a morphological analyzer as in Non-Patent Document 1, when the correct nickname is “U-chan”, the possibility that “-chan” is erroneously extracted as the nickname is reduced. be able to.

  As described above, in the nickname estimation apparatus according to the fourth embodiment, nickname candidates can be extracted from external data such as the Web, so that more appropriate nickname candidates can be selected. For example, it is difficult to estimate a nickname that does not include any characters in the name by using only the nickname generation rule, but it is possible to extract such a nickname by referring to external data.

  Next, the hardware configuration of the nickname estimation apparatus according to the first to fourth embodiments will be described with reference to FIG. FIG. 19 is a hardware configuration diagram of the nickname estimation apparatus according to the first to fourth embodiments.

  The nickname estimation apparatus according to the first to fourth embodiments communicates with a control device such as a CPU (Central Processing Unit) 51 and a storage device such as a ROM (Read Only Memory) 52 and a RAM 53 connected to a network. A communication I / F 54, an external storage device such as an HDD and a CD (Compact Disc) drive device, a display device such as a display device, an input device such as a keyboard and a mouse, and a bus 61 that connects each unit. The hardware configuration uses a normal computer.

  The nickname estimation program executed by the nickname estimation apparatus according to the first to fourth embodiments is a file in an installable format or an executable format, and is a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD). ), A CD-R (Compact Disk Recordable), a DVD (Digital Versatile Disk), and the like.

  Further, the nickname estimation program executed by the nickname estimation apparatus according to the first to fourth embodiments is stored on a computer connected to a network such as the Internet, and is provided by being downloaded via the network. It may be configured. Moreover, you may comprise so that the nickname estimation program performed with the nickname estimation apparatus concerning 1st-4th embodiment may be provided or distributed via networks, such as the internet.

  Moreover, you may comprise so that the nickname estimation program of 1st-4th embodiment may be previously incorporated in ROM etc. and provided.

  The nickname estimation program executed by the nickname estimation apparatus according to the first to fourth embodiments includes the above-described units (name input unit, candidate generation unit, output unit, selection unit, learning data input unit, learning unit, candidate As the actual hardware, the CPU 51 (processor) reads out the nickname estimation program from the storage medium and executes it to load the respective units onto the main storage device. Each unit is generated on the main memory.

  As described above, the apparatus, method, and program according to the present invention are suitable for an information search apparatus, a voice recognition apparatus, a voice interactive apparatus, and the like that can be processed not only by names but also by nicknames.

It is a block diagram of the nickname estimation apparatus concerning 1st Embodiment. It is a figure which shows an example of a nickname production | generation rule. It is a figure which shows an example of the input name data. It is a figure which shows the example which converts a formal name into a symbol. It is a figure which shows the example which converts the hiragana notation into a symbol. It is a figure which shows the example which converts a katakana notation into a symbol. It is a flowchart which shows the whole flow of the nickname estimation process in 1st Embodiment. It is a block diagram of the nickname estimation apparatus concerning 2nd Embodiment. It is a flowchart which shows the whole flow of the nickname estimation process in 2nd Embodiment. It is a block diagram of the nickname estimation apparatus concerning 3rd Embodiment. It is a flowchart which shows the whole flow of the learning process in 3rd Embodiment. It is a figure which shows an example of the produced | generated nickname production | generation rule. It is a figure which shows the usage example of the learned nickname production | generation rule. It is a block diagram of the nickname estimation apparatus concerning 4th Embodiment. It is a flowchart which shows the whole flow of the nickname estimation process in 4th Embodiment. It is a flowchart which shows the whole flow of the nickname candidate extraction process in 4th Embodiment. It is a figure which shows an example of a suffix. It is a schematic diagram which shows an example of the acquisition method of the number of types. It is a hardware block diagram of the nickname estimation apparatus concerning the 1st-4th embodiment.

Explanation of symbols

51 CPU
52 ROM
53 RAM
54 Communication I / F
61 bus 100 nickname estimation device 101 name input unit 102 candidate generation unit 103 output unit 121 rule storage unit 201 nickname candidate 301 person name 302 katakana notation 800 nickname estimation device 804 selection unit 822 document storage unit 1000 nickname estimation device 1005 learning data input unit 1006 Learning unit 1201 Official name 1202, 1212 Nickname 1203, 1213, 1223 Nickname generation rule 1211 Hiragana notation 1301 Official name 1302 Hiragana notation 1303 Katakana notation 1311-1314 Nickname candidate 1400 Nickname estimation device 1404 Selection unit 1407 Candidate extraction unit 1701 Character string 1711-1 1718 Suffix 1801, 1802 characters

Claims (10)

A nickname estimation device that estimates a nickname of a name from a name,
Among the characters included in the name, a rule storage unit that stores position information indicating a position of a character to be included in the nickname candidate and a generation rule for the nickname candidate including a predetermined additional character string;
A name input unit for inputting the name;
Among the characters included in the input name, the character at the position represented by the position information of the generation rule is acquired, and the nickname candidate obtained by combining the acquired character and the additional character string of the generation rule A generating unit for generating
An output unit for outputting the generated nickname candidates;
A nickname estimation device characterized by comprising: A learning data input unit for inputting learning data in which the name and a nickname already known for the name are associated;
A common character included in common in the name of the learning data and the nickname of the learning data is acquired, position information representing a position of the common character is generated from the name of the learning data, and the learning data A learning unit that generates a character string in which the common character is deleted from the nickname, and learns the generated position information and the generated character string as a generation rule for the nickname candidate;
The nickname estimation apparatus according to claim 1. Determining whether or not the nickname candidate is included in a predetermined first document and further including a selection unit that selects the nickname candidate included in the first document;
The nickname estimation apparatus according to claim 1. The selection unit further searches the second document including the selected candidate for the nickname among predetermined second documents, and is determined in advance in descending order of the number of the second documents searched. Selecting a number of said nickname candidates;
The nickname estimation apparatus according to claim 3. The selection unit determines whether the nickname candidate is included in the first document including the input name among the first documents, and the nickname included in the first document. Selecting candidates for
The nickname estimation apparatus according to claim 3. The selection unit obtains a character string having a predetermined number of characters before and after the name from the first document including the inputted name, and the nickname for the obtained character string Determining whether or not a candidate is included, and selecting the nickname candidate included in the acquired character string;
The nickname estimation apparatus according to claim 5. The third document including the inputted name is searched for from among the predetermined third documents, and at least one character having a predetermined number of characters before and after the name of the searched third document. Further comprising an extraction unit for acquiring a column and extracting the nickname candidate from the acquired character string;
The nickname estimation apparatus according to claim 1. A selection unit for selecting the nickname candidate designated by the user from the outputted nickname candidates;
The nickname estimation apparatus according to claim 1. A nickname estimation method executed by a nickname estimation device that estimates a nickname of a name from a name,
The nickname estimation apparatus stores position information indicating a position of a character to be included in the nickname candidate among characters included in the name, and a nickname candidate generation rule including a predetermined additional character string. A rule storage unit,
A name input step in which the name input unit inputs the name;
The generation unit acquires the character at the position represented by the position information of the generation rule among the characters included in the input name, and combines the acquired character and the additional character string of the generation rule Generating the nickname candidate; and
An output unit for outputting the generated nickname candidate;
A nickname estimation method characterized by comprising: A nickname estimation program executed by a nickname estimation device that estimates a nickname of the name from a name,
The nickname estimation apparatus stores position information indicating a position of a character to be included in the nickname candidate among characters included in the name, and a nickname candidate generation rule including a predetermined additional character string. A rule storage unit,
A name input procedure for inputting the name;
Among the characters included in the input name, the character at the position represented by the position information of the generation rule is acquired, and the nickname candidate obtained by combining the acquired character and the additional character string of the generation rule A generation procedure for generating
An output procedure for outputting the generated nickname candidates;
A nickname estimation program for causing the nickname estimation apparatus to execute the program.

Images