JP2012008239A - User dictionary registration device, user dictionary registration method and user dictionary registration program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable accurate accent-linked attributes to be readily assigned.SOLUTION: By a user dictionary registration method, inputting of information on a registered word including notation, pronunciation and accentuation is accepted; a word to be verified is linked with the registered word to generate a linked word to be verified; the linked word to be verified is presented to the user, from whom information on the accentuation of the linked word to be verified is accepted, and the accentuation of the linked word to be verified is determined; on the basis of the determined accentuation, the accentuation of the registered word before structuring the linked word to be verified and that of the registered word after structuring the linked word to be verified are compared, and the accent-linked attribute of the registered word is presumed from the change in accentuation; the obtained information on the accent-linked attribute is assigned to the inputted registered word, along with the information on notation, pronunciation and accentuation, and the registered word is registered in a user dictionary.

Description

  The present invention relates to speech synthesis, and in particular, to providing an accent combining attribute when registering a user dictionary.

  The “accent combination attribute” is an attribute obtained by classifying the change in accent of each word from before combination to after combination when a plurality of words are combined to form one combined word. Here, the difference between “accent” and “accent coupling attribute” in the standard word will be briefly described.

  For example, as shown in FIG. 1 (a), the words "voice" and "synthetic", when used alone, have the sounds of "on" (high, low, low) and "on" (low, high, high). Spoken at high and low. At this time, the portion where the sound changes from high to low is called an accent nucleus, and the type of accent is classified according to the number of beats from the beginning of the word. Therefore, in the case of “speech”, since there is an accent nucleus at the first beat, it is called a type 1 accent. A word that does not have an accent nucleus, such as “composite”, is called a 0-type accent. The definition of the pitch of a word like 0 type and 1 type is defined as “accent”.

  Next, consider a combined word in which a plurality of words are combined. Combined words have accents as well as individual words. The combined word “speech synthesis”, which is a combination of the words “speech” and “synthetic”, is uttered by the pitch of the sound of ↑ onsego ↓ -sei (low high high high high low low) in the standard language, and the accent core is Since it is in the 5th beat, it is classified as a 5-type accent.

  Here, as shown in FIG. 1B, when attention is paid to each of “voice” and “synthesis” of “speech synthesis” uttered as a combined word, the accent changes in the case of the original single utterance. I understand that. That is, the “speech” portion is of type 1 alone, but is of type 0 in combined words. The “synthetic” portion is of type 0 for single utterances but type 1 for combined words.

  Therefore, in consideration of the combination of words, it can be said that the word “speech” has an attribute of “combining with the following word and changing the accent to 0 type”. In addition, it can be said that the word “composite” has an attribute of “combining with the previous word and changing the accent to type 1”. These are “accent coupling attributes”, which are attributes obtained by classifying accent changes at the time of word coupling.

  These accent coupling attributes are different for each word, and are preferably registered in a basic language dictionary used for reading text input in speech synthesis and analyzing accents. In Japanese Patent Laid-Open No. 2001-075585, an “accent coupling attribute” is not given to each word, but “accent coupling attribute” (referred to as an accent class in this document) is obtained from word attribute information such as a word part of speech or a utilization form. A method for creating a rule for automatically and statistically determining “.

  This method has an advantage of saving the trouble of providing an “accent combination attribute” for each word registered in the basic language dictionary. However, the method for statistically obtaining a rule cannot deal with exceptional words that do not follow the rule because the rule is affected by a large amount of data. In order to give an accurate accent, it is necessary to set an accent combining attribute for each word.

  By the way, in speech synthesis, a user dictionary registration tool that can be used by the user is prepared. In speech synthesis, it is necessary to analyze kanji readings and accents. For this reason, a basic language dictionary in which words frequently appearing in Japanese are registered is prepared. However, in reality, not all words that appear in Japanese can be registered in the basic language dictionary, and there are also new words that appear, so many words that cannot be covered by the basic dictionary also appear. As a supplement to this, a user language dictionary is provided in which the user can register notation and readings of words that cannot be read by himself / herself.

JP 2001-077555 A

  There are various variations in the “accent coupling attribute”, and the accent coupling attribute varies depending on the word. Furthermore, as explained above, it is necessary to have specialized knowledge about whether words are combined, accent type, and changes in accent type before and after combining. It is difficult to understand "joint attributes".

  Therefore, when registering the user dictionary, only the word notation, reading, accent, part of speech information is input via the user interface, and the result is given an accent combining attribute roughly classified by the part of speech information, It is common for a user to determine only a limited number of attributes, and an accurate accent combination attribute cannot be assigned. As a result, there was a problem that the accent was wrong during synthesis.

  Therefore, in the case of pursuing reading accuracy, it is necessary to give an accurate “accent combination attribute” for each word, and a general user who does not know a complicated “accent combination attribute” can provide an accurate accent combination attribute. It was desired.

  Therefore, an object of the present invention is to make it possible to easily assign an accurate accent coupling attribute.

  In order to achieve the above object, the present invention relates to a registered word information input unit that accepts input of information including at least notation, reading, and accent related to a registered word, and a verification word by combining the verification word with the registered word. A verification combined word generation unit that generates a word; and the verification combined word generated by the verification combined word generation unit is presented to the user, receives information on the accent of the verification combined word from the user, and sets the accent of the verification combined word Based on the verification combined word accent determination unit to be determined and the accent determined by the verification combined word accent determination unit, the accent after the registration word and the verification combined word before configuring the verification combined word Accent that compares the accent of the registered word and estimates the accent combining attribute of the registered word from the change of the accent Accompanied by the combined attribute estimation unit and the registered word input by the registered word information input unit, information on the accent coupling attribute obtained from the accent coupling attribute estimation unit is added to information including at least notation, reading, and accent. And a user dictionary registration unit for registering in the user dictionary.

  According to said structure, even if it is not the expert who is familiar with the knowledge of the accent coupling | bonding attribute, an accurate accent coupling | bonding attribute can be provided easily.

Diagram explaining accents and accent join attributes 1 is a block diagram showing an example of the overall configuration of a speech synthesizer according to Embodiment 1 of the present invention. The flowchart which shows an example of operation | movement of the speech synthesizer which concerns on Embodiment 1 of this invention. The figure which shows an example of the user interface The figure which shows an example of verification information DB which concerns on Embodiment 1 of this invention. The block diagram which shows an example of a structure of the verification combined word production | generation part in Embodiment 1 of this invention. The flowchart which shows an example of operation | movement of the verification combined word production | generation part in Embodiment 1 of this invention. The block diagram which shows an example of a structure of a verification joint word accent determination part. Flow chart showing an example of the operation of the verification combined word accent determination unit The figure which shows an example of the user interface The block diagram which shows an example of a structure of an accent joint attribute estimation part Flow chart showing an example of the operation of the accent combination attribute estimation unit The figure which shows an example of verification information DB which concerns on Embodiment 2 of this invention. The block diagram which shows an example of a structure of the verification combined word production | generation part in Embodiment 2 of this invention. The flowchart which shows an example of operation | movement of the verification combined word production | generation part in Embodiment 2 of this invention. The block diagram which shows an example of a structure of the verification combined word production | generation part in Embodiment 3 of this invention.

[Embodiment 1]
FIG. 2 is a block diagram showing an example of the configuration of the user dictionary registration apparatus 100 according to Embodiment 1 of the present invention. In FIG. 2, the user dictionary registration device 100 includes a registered word information input unit 110, a verification combined word generation unit 120, a verification information DB 130 (verification information storage unit), a verification combined word accent determination unit 140, an accent combined attribute estimation unit 150, And a user dictionary registration unit 160. FIG. 3 is a flowchart showing an example of the operation of the user dictionary registration apparatus 100 according to the first embodiment of the present invention. Hereinafter, description will be given according to the block diagram of FIG. 2 and the flowchart of FIG.

  The registered word information input unit 110 accepts input of the word notation, reading, accent, and part of speech for the word that the user wants to register (step S301). An example of the user interface at that time is shown in FIG. The registered word information input unit 110 accepts input of notation, reading, and accent from boxes for specifying notation, reading, and accent, respectively (reading and accent are specified in the same box). The accent can be specified by adding “′” to the position of the accent nucleus of the reading expressed in katakana, for example. The method of designating the accent is not limited to this, and other methods of giving may be used. The part of speech is selected from the list. The part of speech is also an example of word attribute information used when generating a verification combined word described later.

  The verification combined word generation unit 120 refers to the verification information DB 130 (verification information storage unit), and generates a verification combined word for verifying the accent combined attribute of the registered word (step S302). The verification combined word is a word obtained by combining the registered word and the verification word.

  For example, the registered word information input unit 110 can accept input of information including the word attribute of the registered word in addition to the notation, reading, and accent related to the registered word. The verification information DB 130 stores verification words in association with word attributes as verification information. The verification combined word generation unit 120 extracts the verification word associated with the verification information of the word attribute that matches the word attribute of the registration word received by the registration word information input unit 110 from the verification information DB 130 and registers it. By combining with the word, a verification combined word can be generated. The method of determining the verification word is not limited to the above example. For example, a predetermined fixed word may be used as the verification word, or a verification word that is natural Japanese may be selected according to the reading of the registered word. Further, the verification information DB 130 does not need to be provided in the user dictionary registration device 100.

  The verification combined word accent determination unit 140 presents the verification combined word generated by the verification combined word generation unit 120 to the user. The verification combined word accent determination part 140 receives the input of the accent information regarding a verification combined word from a user, and determines the accent of a verification combined word (step S303).

  The accent combination attribute estimation unit 150 compares the accent of the registered word before forming the verification combined word with the accent of the registered word portion after forming the verification combined word. That is, the accent combining attribute estimation unit 150 compares the accent of the registered word before combining with the accent of the registered word after combining, and extracts the change in the accent before and after combining the registered words as the accent change of the registered word.

  The accent combination attribute estimation unit 150 compares the accent change of the registered word obtained from the comparison result with the accent change indicated by the accent combination attribute registered in the verification information DB 130, and determines the accent change of the registered word. The accent combination attribute indicating the matching accent change is estimated as the accent combination attribute of the registered word (step S304).

  The accent combination attribute is registered in advance in the verification information DB 130 as data representing an accent change rule (rule) before and after composing a verification combination word, for example, for each word attribute. For example, the verification information DB 130 sets, as an accent coupling attribute, an accent coupling attribute that represents a method of changing accents by combining words and a verification word that causes the accent coupling, and associates them with the word attribute. Stored. The accent combination attribute estimation unit 150 uses the verification combination word obtained by combining the verification word corresponding to the word attribute of the registration word and the registration word, and determines the accent combination attribute of the registration word from the change in the accent of the registration word before and after the combination. presume. For example, the accent combination attribute can be estimated by searching the verification information DB 130 for data that matches the word attribute of the registered word and the registered word accent change before and after the combination.

  Specifically, the accent combination attribute estimation unit 150 includes a change from the accent of the registered word before configuring the verification combined word to the accent of the registered word part after configuring the verification combined word, and the verification information storage unit It is possible to compare the accent combination attribute that matches the word attribute of the registered word to be stored. The accent combination attribute estimation unit 150 can estimate the accent combination attribute corresponding to the accent change as the accent combination attribute of the registered word.

  Further, the verification information DB 130 may store only word attributes and corresponding verification information. In this case, the accent combination attribute estimation unit 150 compares the accent of the registered word before forming the verification combined word with the accent of the registered word part after forming the verification combined word, and calculates the change in accent before and after the combination. It is estimated as an accent combination attribute as it is. With such a configuration, the size of the verification information storage unit can be reduced, and the search process in the estimation process can be reduced. Further, by estimating the change in accent before and after the combination as an accent combination attribute, it is possible to deal with an accent combination attribute that was not anticipated in advance.

  The user dictionary registration unit 160 uses the accent coupling attribute estimated by the accent coupling attribute estimation unit 150 for the word received by the registered word information input unit 110 and registered as information on the registered word along with the notation, reading, and accent. The user dictionary is registered in the user dictionary 170 (step S305).

  FIG. 5 is an example of the verification information DB 130 according to the first embodiment. In FIG. 5, registered in the verification information DB 130 are “word attributes” of registered words, “accent coupling attributes” that can be taken by the words of the attributes, and “verification information” for verifying the accent coupling attributes. ing.

  The “word attribute” is an example of attribute information for classifying accent coupling attributes. In the example shown in FIG. 5, the “word attributes” are classified according to the part of speech and the accent type, such as those classified as other than the non-variable noun 0 type (note), other than the proper noun 0 type, etc. . (Note: Sa-variant nouns refer to nouns connected to sa-line inflection verbs.) Part of speech is not limited to this, for example, common nouns, verbs, verb conjugations, adjectives, adjective verbs, case particles. In addition to grammatical classifications such as connection particles, gender, name, country name, place name, etc., which further classify parts of speech may be used as necessary.

  The “accent coupling attribute” indicates an accent coupling attribute that can be taken by the word attribute. ”And“ (2) Attributes that may change to type 1 by combining accents with the immediately preceding word ”. Note that (1) an attribute that does not accent join with the immediately preceding word is an attribute that does not change the accent type of the registered word before and after combining even if the registered word is joined with the immediately preceding word. (2) The attribute that may be combined with the immediately preceding word to change to type 1 is that the registered word is combined with the immediately preceding word, so that the registered word is from an accent type other than type 1 before and after the combination. It is an attribute that may change to type 1 accent type.

  In the “verification information”, information for verifying the accent combining attribute that each word attribute can take is written, and in the first embodiment, the verification word includes a verification word to be combined with the registered word. For example, if the word attribute of the registered word is other than the sub-noun 0 type and its notation is ΔΔ, in FIG. 5, the verification word is “new”, and the verification combined word combines the verification word and the registered word. It becomes the word “new ΔΔ”.

  FIG. 6 is a block diagram illustrating an example of the configuration of the verification combined word generation unit 120 according to the first embodiment. In FIG. 6, the verification combined word generation unit 120 includes a word attribute extraction unit 121, a verification word acquisition unit 122, and a word combination unit 123. FIG. 7 is a flowchart showing an example of the operation of the verification combined word generation unit 120. Hereinafter, description will be given according to the block diagram of FIG. 6 and the flowchart of FIG.

  The word attribute extraction unit 121 extracts, from the verification information DB 130, the word attribute of the word received by the registered word information input unit 110 (step S701), that is, the word attribute having the same part of speech and accent type in this embodiment. (Step S702). For example, if the registered word is “download sale”, the reading accent is “downloaded hanbai”, and the part of speech is specified as “sa variable noun”, the word attribute extraction unit 121 reads “other than 0 type of sa variable noun”. The word attribute (that is, the accent type is other than 0 type) is extracted.

  The verification word acquisition unit 122 acquires a verification word corresponding to the word attribute extracted by the word attribute extraction unit 121. In the example of the verification information DB 130 in FIG. 4, verification words “new” and “shinki” (that is, accent type 1 type) are registered as verification information corresponding to the word attribute “other than 0 type of sa strange noun”. Therefore, “new” is acquired as a verification word as it is (step S703).

  Since the verification word is registered as the left verification word, the word combining unit 123 combines the verification word “new” with the registration word “download sales” to generate a verification combined word “new download sales”. (Step S704). Then, the process proceeds to the process of determining the verification combined word accent (step S705).

  FIG. 8 is a block diagram illustrating an example of the configuration of the verification combined word accent determination unit 140. In FIG. 8, the verification combined word accent determination unit 140 includes a presentation unit 141, an input reception unit 142, and an accent determination unit 143. FIG. 9 is a flowchart showing an example of the operation of the verification combined word accent determination unit 140. Hereinafter, description will be given according to the block diagram of FIG. 8 and the flowchart of FIG.

  The presentation unit 141 displays the verification combined word generated by the verification combined word generation unit 120 (step S901), for example, as shown in FIG. Step S902), prompting the user to input the verification combined word accent.

  The input receiving unit 142 receives the input of the accent of the verification combined word from the user with respect to the notation of the verification combined word presented by the presenting unit 141 (step S903).

  The accent determination unit 143 obtains the position of the accent nucleus from the accent of the verification combined word input by the user, and determines the accent type of the verification combined word (step S904). As shown in FIG. 10A, when the user inputs an accent “Shinki_Download Hanby” to the notation of the verification combined word “new download sales”, the position of the accent nucleus (“′”). From the position of “1”, “Shinki” is determined to be type 1 (ie, accent type is 1 type), and “Download Hanby” is determined to be type 5 (ie, accent type is other than 0 type) (“_” is Represents an accent phrase boundary, and the position of the accent nucleus of the download standby is counted from the beginning of the accent phrase = accent boundary).

  In addition, the presentation of the verification combined word and the determination of the accent in the presentation unit 141, the input reception unit 142, and the accent determination unit 143 are not limited to the above-described methods. For example, the presentation unit 141 refers to the verification information DB 130 together with the notation of the verification combined word, and presents a plurality of accent candidates according to the possible accent combination attributes to the user. The candidates may be selected and determined. Further, the user presenting unit 141 refers to the verification information DB 130 for the verification combined word generated by the verification combined word generation unit, generates a synthesized speech of accents according to possible accent coupling attributes, and reproduces the synthesized speech. A screen for this may be presented to the user as shown in FIG. In this case, the user plays back the presented verification combined word (synthesized speech) and presses the stop button to view it, and the input accepting unit 142 accepts an input from the user indicating whether the reading (accent of the synthesized speech) is correct or incorrect. .

  When the user inputs that the accent is correct, the accent determination unit 143 determines the first generated accent as the verification combined word accent. If the user inputs an error, the accent determination unit 143 uses another accent combination attribute to generate an accent of the verification combined word, presents it again to the user, and determines the accent.

  FIG. 11 is a block diagram illustrating an example of the configuration of the accent combination attribute estimation unit 150. In FIG. 11, the accent connection attribute estimation unit 150 includes a registered word accent extraction unit 151 and an accent connection attribute estimation unit 153. FIG. 12 is a flowchart showing an example of the operation of the accent joint attribute estimation unit 150. Hereinafter, description will be given according to the block diagram of FIG. 11 and the flowchart of FIG.

  After the verification combined word accent is determined (step S1201), the registered word accent extraction unit 151, in this embodiment, is the accent type of the registered word input by the registered word information input unit 110, that is, “download type”. Accent type indicated by 'Nmby' (that is, 7 type = accent type is other than 0 type) is acquired (step S1202).

  The accent combination attribute estimation unit 153 is an accent combination attribute when the change from the accent type of the registered word before combination to the accent type of the registered word after combination determined by the verification combined word accent determination unit matches. The accent combination attribute associated with the word attribute corresponding to the word is acquired from the verification information DB 130, and is estimated as the accent combination attribute of the registered word (step S1203). In the present embodiment, the word attribute “other than sao noun 0 type” is matched, and the verification combined word accent type “Shinki Download Hanbai” is combined according to “(1) attribute that does not accent combine with the previous word”. Since the accent type indicated by the “Download Humbai” portion of the registered word matches the accent “Download Humbai” of the registered word alone, the accent combination attribute of the registered word “Download Sales” is “(1) the word just before It is presumed that the attribute is not an accent combination.

  Then, the user dictionary registration unit 160 gives the accent combination attribute estimated by the accent combination attribute estimation unit 150 as information of the registered word for the word that the user accepts from the registered word information input unit 110 and wants to register. The user dictionary is registered in the user dictionary 170 together with the reading and accent (step S1204). In the present embodiment, the user dictionary registration unit 160 adds the information “attribute that does not accent combine with the immediately preceding word” as information of the accent combination attribute of the registered word “download sale” and registers it in the user dictionary 170.

  As described above, according to the present embodiment, when a user registers a word in the user dictionary, notation, reading, accent, and part-of-speech information are input, and further, accent information for the presented verification combined word is displayed. If it is input, it is possible to easily add the information of the accent coupling attribute.

[Embodiment 2]
In the present embodiment, in the “verification information” column of the verification information DB 130, not the verification word itself but a condition to be satisfied by the verification word is registered, and the verification word acquisition unit 124 reads from the basic dictionary 180 and the user dictionary 170. The verification word is searched based on the condition to be satisfied by the verification word.

  FIG. 13 is an example of the verification information DB 130 in the second embodiment. The “word attribute” and “accent combination attribute” are the same as those in the verification information DB 130 in the first embodiment, but the “verification information” includes, for example, “on the left side of the registered word, the accent of the single utterance is other than 0 type” A condition that the verification word should satisfy is registered such as “Accent type of and having a record of combined accents”. The condition to be satisfied by the verification word is, for example, a flag indicating whether or not the registered word is on the left side, a flag indicating whether or not the accent type is other than 0 type, and whether or not there is an accent combination. It is represented by data of a set of flags.

  FIG. 14 is a block diagram showing an example of the configuration of the verification combined word generation unit 120 in the user dictionary registration 100 according to Embodiment 2 of the present invention. In this embodiment, as compared with the configuration of the first embodiment, the verification combined word generation unit 120 includes a verification condition acquisition unit 126 and a verification word search unit 127 instead of the verification word acquisition unit 122. The verification word acquisition unit 124 is included, and the processing of the verification combined word generation unit 120 is different. Components other than the verification combined word generation unit 120 that are the same as those in the first embodiment are denoted by the same reference numerals, and descriptions of processes that are the same as those in the first exemplary embodiment other than the verification combined word generation unit 120 are omitted.

  The verification condition acquisition unit 126 acquires a verification condition corresponding to the word attribute extracted by the word attribute extraction unit 121 with reference to the verification information DB 130.

  The verification word search unit 127 searches the basic dictionary 180 and the user dictionary 170 for verification words that match the verification conditions acquired by the verification condition acquisition unit 126, and outputs them to the word combination unit 123.

  FIG. 15 is a flowchart showing an example of the operation of the verification combined word generation unit 120 in the user dictionary registration apparatus 100 according to Embodiment 2 of the present invention. Hereinafter, description will be given with reference to the flowchart of FIG.

  Here, the registered word information input unit 110 inputs the notation “download sales”, the reading / accent as “download hammbai”, and the part of speech as “sa variable noun” via the interface shown in FIG. Is accepted (step S1501).

  Similarly to the first embodiment, the word attribute extraction unit 121 obtains matching word attributes from the verification information DB 130 shown in FIG. 13 based on the part of speech and accent type of the registered word input by the registered word information input unit 110. Extract. Here, the word attribute extraction unit 121 obtains the position of the accent nucleus (the position of “′”) from “download road” input by the registered word information input unit 110 and obtains the sixth beat = 6 type. Based on this and the part-of-speech information input, the word attribute “other than the type 1 of the sabot noun” is extracted (step S1502).

  The verification condition acquisition unit 126 acquires a verification condition that matches the word attribute extracted by the word attribute extraction unit 121. In the present embodiment, since it is assumed that the word attribute is “other than type 1 of the Soma noun”, referring to the verification information DB 130 shown in FIG. The verification condition “is an accent type other than type 0 and has a combined record” is acquired (step S1503). As described above, at the time of implementation, a set of various flags representing conditions is acquired.

  The verification word search unit 127, when combined with an arbitrary word having a word attribute of a registered word to form a combined word, searches the basic dictionary 180 for a verification word that matches the verification condition acquired by the verification condition acquisition unit 126. And it searches from the user dictionary 170 (step S1504).

  In the present embodiment, the word attribute of the registered word is “other than the type 1 of the sabot noun”, the verification condition is “an accent type other than the 0 type on the left side of the registered word and the accent of the single utterance, and has a record of combined accents. Therefore, a word that satisfies the following conditions is searched for as a verification word.

Construct a combined word in the form of “XX (verification word that satisfies the verification condition) + Δ △ (word having the word attribute of the registered word)”
(1) XX is a verification word (= "on the left side of the registered word"), and in the case of a single utterance, the accent type is "other than 0 type".
(2) ΔΔ is a word having a word attribute of a registered word, and has an accent type of “other than type 1 of the Sabina noun” by a single utterance.
(3) There is a combined word (◯◯ + ΔΔ) between the verification word and the registered word in which accent coupling occurs.

  The search process starts by searching the basic dictionary 180 for words that satisfy the accent condition of (1). For example, if the word “new: sign (accent type is type 1)” is registered in the basic dictionary 180, “new” is “accent type other than type 0” (1). First, it becomes a verification word candidate.

  Next, the basic dictionary 180 and the user dictionary 170 are searched for those satisfying the condition (2). That is, from the basic dictionary 180 and the user dictionary 170, a combined word of “new (verification word candidate that may satisfy the verification condition) + ΔΔ (word having a registered word attribute)” A search is performed as to whether or not there is a part that satisfies “other than savariant type 1” when the part of ΔΔ is a single utterance. Here, it is assumed that the words “new entry: synxon's new” and “new listing: shiinki jojo” are found in the basic dictionary 180 or the user dictionary 170.

  Finally, a candidate that satisfies the condition (3) is searched from the found candidates. In this case, looking at the accent change of the “new” part before and after the combination, in the case of “new entry”, the accent has changed and the accent combination has occurred. In the case of “new listing”, the accent is maintained before and after the coupling, and no accent coupling occurs. As the verification condition of (3), the condition “having a record of combined accents” is required, so that “new entry” satisfies this condition. From the above, it can be seen that “new” matches all the verification conditions (1) to (3) as the verification word, and the word “new” is output as the search result.

  Lastly, the word combining unit 123 generates a verification combined word according to the verification condition. Here, since the verification condition includes a condition “on the left side of the registered word”, the verification word “new” and the registered word “download sales” are combined to generate a verification combined word “new download sales” ( Step S1505). Then, the process proceeds to the process of determining the verification combined word accent (step S1506).

  As described above, according to the present embodiment, since verification words can be extracted from words registered in the basic dictionary or the user dictionary, verification words that are more natural as Japanese can be used, and can be easily and accurately Accent binding attribute can be added.

[Embodiment 3]
In the present embodiment, the verification word search unit 127 constituting the verification word acquisition unit 125 is configured to be able to search the data of another user's dictionary 190 in addition to the user's user dictionary, and a mechanism that can search for a more optimal verification word It is said.

  FIG. 16 is a block diagram showing an example of the configuration of the verification combined word generation unit 120 in the user dictionary registration 100 according to Embodiment 3 of the present invention. The configuration of the second embodiment is different in that the verification word search unit 127 further includes an external dictionary search unit 128 and an internal dictionary search unit 129. The same components as those in the first and second embodiments are denoted by the same reference numerals, and the description of the processing by the same components as those in the first and second embodiments is omitted.

  The verification word search unit 127 searches for a verification word that matches the verification condition acquired by the verification condition acquisition unit 126 from the user dictionary 190 of another person in addition to the basic dictionary 180 and the user dictionary 170, Output to the combining unit 123.

  The internal dictionary search unit 129 searches the basic dictionary 180 and the user dictionary 180 for a verification word.

  The external dictionary search unit 128 is connected to the verification word search unit 127a of the other person's synthesis device, and also searches the verification word search unit 127a of the other person's synthesis device to search for the verification word from the other person's user dictionary 190. .

  In addition, the verification word search part 127 is not restricted to the said structure. As another configuration, individual user dictionaries may be collectively stored in a server, and an external dictionary search unit may search for another user's user dictionary on the server.

  As described above, according to the present embodiment, the data amount of verification words increases, so that a more optimal verification word can be used and an accurate accent combination attribute can be easily provided.

  The configuration described in the above embodiment merely shows a specific example, and does not limit the technical scope of the present invention. Any configuration can be employed within the scope of the effects of the present invention.

  In the embodiment of the present invention, a software program for realizing the above-described embodiment (in the embodiment, a program corresponding to the flowchart shown in FIG. 2) is supplied to the apparatus, and a computer of the apparatus is supplied. Including the case where it is achieved by reading and executing the program. Therefore, in order to realize the functional processing described in this embodiment by a computer, the program itself installed in the computer is also an embodiment of the present invention. That is, it is realized by the function block shown in FIGS. 2, 6, 8, 11, 14, and 16 of the present invention and the CPU executing the program in cooperation. A medium recording a program for realizing the functional processing of the present invention is also included in one aspect of the embodiment.

100 user dictionary registration apparatus 110 registered word information input unit 120 verification combined word generation unit 121 word attribute extraction unit 122 verification word acquisition unit 123 word combination unit 124 verification word acquisition unit 125 verification word acquisition unit 126 verification condition acquisition unit 127 verification word search Unit 128 external dictionary search unit 129 internal dictionary search unit 130 verification information DB (verification information storage unit)
140 Verification combined word accent determination unit 141 Presentation unit 142 Input reception unit 143 Accent determination unit 150 Accent combination attribute estimation unit 151 Registered word accent extraction unit 153 Connection attribute estimation unit 160 User dictionary registration unit 170 User dictionary 180 Basic dictionary 190 Other user dictionary

Claims (8)

A registered word information input unit that accepts input of information including at least notation, reading, and accent on the registered word;
A verification combined word generation unit that combines a verification word with the registered word to generate a verification combined word;
Presenting the verification combined word generated by the verification combined word generation unit to the user, accepting information on the accent of the verification combined word from the user, and determining the accent of the verification combined word;
Based on the accent determined by the verification combined word accent determination unit, the accent of the registered word before configuring the verification combined word and the accent of the registered word after configuring the verification combined word, From an accent change, an accent joint attribute estimation unit that estimates an accent joint attribute of the registered word;
The information about the registered word input in the registered word information input unit is added to the information including at least the notation, the reading, and the accent, and the information of the accent coupling attribute obtained from the accent coupling attribute estimation unit is added and registered in the user dictionary. A user dictionary registration device comprising: a user dictionary registration unit that performs: A verification information storage unit that stores verification information necessary to generate a verification combined word for verifying the accent combined attribute of the registered word;
The verification combined word generation unit obtains a verification word based on verification information stored in the verification information storage unit, and verification for verifying an accent combined attribute of the registered word from the registered word and the verification word The user dictionary registration device according to claim 1, which generates a combined word. The registered word information input unit accepts input of information including a word attribute of the registered word in addition to the notation, reading, and accent related to the registered word,
The verification information storage unit stores a verification word in association with a word attribute as the verification information,
The verification combined word generation unit extracts a verification word corresponding to verification information that matches the word attribute received by the registered word information input unit from the verification information storage unit, and generates the verification combined word The user dictionary registration device according to claim 1 or 2. The verification information storage unit stores, as the verification information, an accent combination attribute that represents a change in accent when a word is combined with another word in association with the word attribute,
The accent joint attribute estimation unit is an accent joint attribute associated with a word attribute that matches the word attribute of the registered word in the verification information storage unit, and is a single registered word before constituting the verification combined word The accent combination attribute that matches the accent change to the accent of the registered word part after constituting the verification combined word from the utterance accent is estimated as the accent combined attribute of the registered word. The user dictionary registration device according to any one of the above.   The verification combined word generation unit retrieves the verification word from the basic language dictionary or the user dictionary based on the verification condition stored in the verification information storage unit by searching for a verification word that matches the condition. The user dictionary registration device according to claim 1, which is generated.   The verification combined word generation unit generates the verification word by searching a verification word that matches the condition from another person's user dictionary based on the verification condition stored in the verification information storage unit. The user dictionary registration device according to claim 1. A registered word information input step for accepting input of information including at least notation, reading, and accent on the registered word;
A verification combined word generation step of combining a verification word with the registered word to generate a verification combined word;
A verification combined word accent determination step of presenting the verification combined word generated in the verification combined word generation step to the user, receiving information on the accent of the verification combined word from the user, and determining an accent of the verification combined word;
Based on the accent determined in the verification combined word accent determination step, the accent of the registered word before configuring the verification combined word and the accent of the registered word after configuring the verification combined word, An accent combining attribute estimating step for estimating an accent combining attribute of the registered word from an accent change;
Information on the registered word input in the registered word information input step is added to the information including at least notation, reading, and accent, and the information of the accent connection attribute obtained from the accent connection attribute estimation unit is added and registered in the user dictionary. A user dictionary registration method comprising: a user dictionary registration step. On the computer,
A registered word information input step for accepting input of information including at least notation, reading, and accent on the registered word;
A verification combined word generation step of combining a verification word with the registered word to generate a verification combined word;
A verification combined word accent determination step of presenting the verification combined word generated in the verification combined word generation step to the user, receiving information on the accent of the verification combined word from the user, and determining an accent of the verification combined word;
Based on the accent determined in the verification combined word accent determination step, the accent of the registered word before configuring the verification combined word and the accent of the registered word after configuring the verification combined word, An accent combining attribute estimating step for estimating an accent combining attribute of the registered word from an accent change;
Information on the registered word input in the registered word information input step is added to the information including at least notation, reading, and accent, and the information of the accent connection attribute obtained from the accent connection attribute estimation unit is added and registered in the user dictionary. A user dictionary registration program for executing a user dictionary registration step.

Images