JP2022029869A - Program, device and method for creating dialog scenario with verbs replaced so as to correspond to character attribute - Google Patents

Abstract

To provide a program, dialog scenario creation device and dialog scenario creation method.for creating a dialog scenario with verbs replaced so as to correspond to a character attribute.SOLUTION: A dialog scenario creation device 1 comprises: dialog scenario accumulation means 101 which accumulates dialog scenarios; contribution sentence accumulation means 102 which accumulates a plurality of contribution sentences of a plurality of users; meaning element determination means 11 which determines a meaning element in a spoken verb described in a spoken sentence of the dialog scenario based on a first character attribute, and associates a plurality of synonym verbs having the same meaning element with each other; feature synonym verb selection means 12 which selects the synonym verb for which an appearance frequency is most lopsided among the plurality of synonym verbs in the same meaning element as a feature synonym verb, by using the user's contribution sentence based on a second character attribute; and verb replacement means 13 which replaces the spoken verb described in the spoken sentence of the dialog scenario based on the first character attribute with the feature synonym verb, to convert the spoken sentence into a spoken sentence of the dialog scenario of the second character attribute.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for creating a dialogue scenario of an agent interacting with a user.

Dialogue systems that naturally interact with users using smartphones, tablet terminals, and robots have become widespread. A computer graphics agent appears on the display and interacts with the user by voice or text. The agent can be recognized as one person who can interact with the user, and establishes a dialogue according to the user's situation, hobbies, and emotions.
Agents have unique character attributes (age, gender, hometown, etc.) as well as a person's profile. The user can have a more intimate feeling, for example, by interacting with an agent having a character attribute that he / she likes.

FIG. 1 is an explanatory diagram showing a sense of incongruity that a user feels through dialogue with an agent.

According to FIG. 1 (a), an agent with a character attribute of a woman in her twenties has a dialogue with a user as follows.
User: "What do you do with rice at night?"
Agent: "I'll eat fried rice."
User: "What is the music you listen to often?"
Agent: "Listen to KANA"

On the other hand, according to FIG. 1 (b), a teenage / male character attribute agent interacts with the user in the same manner using the same dialogue scenario as in FIG. 1 (a).
User: "What do you do with rice at night?"
Agent: "I'll eat fried rice."
User: "What is the music you listen to often?"
Agent: "Listen to KANA"
In this case, the user feels uncomfortable to speak, for example, "I'll eat", even though he / she is interacting with an agent having a character attribute of a teenager / male. This is because teenagers and men often make rough utterances.

Therefore, the agent dialogue scenario generally needs to be manually created according to the character attribute. By creating it manually, it is possible to create an agent that feels individuality according to the character attributes. However, there is a cost problem in manually creating dialogue scenarios for various agents.

Conventionally, there is a technique for automatically rewriting each phrase of an utterance sentence according to an agent having a desired character attribute (see, for example, Patent Document 1). According to this technique, a "rewriting rule collection" of utterances created by hand is referred to according to character attributes such as age and gender.
For example, when creating a dialogue scenario with a character attribute of a teenager / woman from the dialogue scenario "I like golf" with a character attribute of a man in his 50s, "na" is rewritten as "yo". This creates "I like golf" as a dialogue scenario for teenage and female character attributes.
In general, golf is a popular sport for men in their 50s. Therefore, there is no sense of discomfort when a male agent in his 50s says, "I like golf." However, it feels strange for a teenage female agent to say, "I like golf." This is because it is not the content spoken by general teenagers and women.

On the other hand, there is a technique of replacing a noun included in an utterance sentence of a dialogue scenario with a noun suitable for a desired character attribute (see, for example, Non-Patent Document 1). According to this technique, a dialogue scenario tailored to a character attribute is created by performing the following steps.
(S1) A character attribute (for example, a woman in her twenties, a man in her fifties, etc.) is specified.
(S2) With respect to the noun included in the utterance sentence of the character in the dialogue scenario, similar candidates of the noun are extracted from Wikipedia (registered trademark).
(S3) From the extracted similar candidates, the posted text of the SNS user having the same character attribute as the specified character attribute is extracted.
(S4) One noun that appears characteristically in the posted sentence of the SNS user is determined.
(S5) The noun included in the utterance sentence of the character in the dialogue scenario is replaced with the noun determined in S4.
For example, the utterance sentence "I often listen to KANA" in the dialogue scenario of the character attribute of a woman in her twenties is rewritten as the utterance sentence "I often listen to HAMADA" in the dialogue scenario of the character attribute of a man in her fifties. be able to.

Japanese Unexamined Patent Publication No. 2013-196070

Shunichi Tahara et al., "Proposal and evaluation of automatic replacement method for nouns that emphasizes character in dialogue scenarios", DEIM 2019, [online], [Search on July 8, 2nd year of Reiwa], Internet <URL: https: //db-event.jpn.org/deim2019/post/papers/2.pdf ＞ Wikipedia: Category "Verb", [online], [Search on July 1, 2nd year of Reiwa], Internet <URL: https://ja.wikipedia.org/wiki/Category:%E5%8B%95%E8% A9% 9E ＞ Weblio Thesaurus, [online], [Search on July 8, 2nd year of Reiwa], Internet <URL: https://thesaurus.weblio.jp/> Associative Thesaurus, [online], [Search on July 8, 2nd year of Reiwa], Internet <URL: https://renso-ruigo.com/word/> Synonyms dictionary, [online], [Search on July 8, 2nd year of Reiwa], Internet <URL: https://dictionary.goo.ne.jp/thsrs/>

However, for example, as in the technique described in Non-Patent Document 1, the user may feel uncomfortable with the spoken sentence of the dialogue scenario only by replacing the noun.
For example, "KANA" is an artist popular with women in their twenties, and "HAMADA" is an artist popular with men in their fifties.
At this time, it is assumed that there is "I was struck by my KANA CD" as an utterance sentence of an agent with a character attribute of a woman in her twenties. It is assumed that the noun included in this utterance is replaced with the utterance sentence of a character for men in their 50s by the technique described in Non-Patent Document 1. Here, "KANA" included in the utterance of a female character in her twenties is replaced with "HAMADA" suitable for a male character in her fifties. Then, it will be replaced with the utterance sentence "My HAMADA CD was packed".
However, the verb "pakuru" in the rewritten utterance sentence is a word often used by young people, and it feels strange as an utterance that characterizes a man in his 50s. That is, since the verb included in the utterance sentence is not replaced with the verb that matches the character attribute, a sense of incongruity remains as a character expression.

On the other hand, the inventors of the present application thought that simply replacing a noun or a phrase in a dialogue scenario would not represent the characteristics of a person based on the character attribute. In fact, I thought that the "verb" that the person based on the character attribute is reminded to actually speak might be the most representative of the characteristics of the person based on the character attribute.

Also, for example, pages on Wikipedia (registered trademark) are dominated by nouns, and there are few articles related to verbs. Therefore, there is also a problem that there are few verb replacement candidates. For example, looking at the category "verbs", there are only 47 articles about verbs. Therefore, it is extremely difficult to replace the verb contained in the utterance sentence of the dialogue scenario with the verb corresponding to the character attribute by simply referring to the verb of Wikipedia.

Therefore, an object of the present invention is to provide a program, a device, and a method for creating a dialogue scenario in which a verb is replaced according to a character attribute.

According to the program of the present invention
Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
Post-sentence storage means for accumulating multiple posts by multiple users for each character attribute,
Meaning element determination means for determining the meaning element in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs with the same meaning element.
Characteristic synonymous verb selection means for selecting a synonymous verb whose appearance frequency is most biased among multiple synonymous verbs in the same meaning element as a characteristic synonymous verb using a user's post based on the second character attribute. When,
Replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic verb, and making the computer function as a verb replacement means to be the utterance sentence of the dialogue scenario of the second character attribute. It is characterized by.

According to other embodiments in the program of the invention
Meaning factor determination means
When multiple meanings are determined for the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute,
Using the synonym corpus, search for multiple synonymous verbs that have the same meaning for each meaning element.
In the user's post based on the first character attribute, the frequency of appearance of each synonymous verb is counted for each meaning element, and
It is also preferable to make the computer function so as to determine the meaning element having the highest frequency of appearance of each synonymous verb.

According to other embodiments in the program of the invention
Meaning factor determination means
Part-of-speech extraction means for extracting nouns and verbs that are dependent on each other from the utterance sentences of the dialogue scenario based on the first character attribute.
A synonym verb search method for searching for synonymous verbs similar to dependent verbs for each meaning element using a synonym corpus,
The co-occurrence frequency of searching the co-occurrence frequency of the dependent noun and each synonymous verb from the user's post based on the first character attribute, and determining the co-occurrence element with the highest co-occurrence frequency of each synonymous verb. It is also preferable to make the computer function as a search means.

According to other embodiments in the program of the invention
Characteristic verb selection means
As shown below, using the Akaike's Information Criterion (AIC), the evaluation value V (P1, P2) of the synonymous verb P2 for the character attribute P1 is calculated, and the evaluation value V (P1, P2). The synonymous verb P2 whose evaluation value V (P1, P2) is greater than or equal to the predetermined threshold or whose evaluation value V (P1, P2) is the largest is the predetermined number.
p11: The number of posts related to the synonymous verb P2 by the user of the character attribute P1.
p12: The number of posted sentences other than the synonymous verb P2 by the user of the character attribute P1.
p21: The number of posts related to the synonymous verb P2 by users other than the character attribute P1.
p22: The number of posted sentences other than the synonymous verb P2 by users other than the character attribute P1.
p11 ＋ p12 ＋ p21 ＋ p22 ＝ Number of posts posted by all users
MLL_IM (P1, P2) = (p11 + p12) log (p11 + p12)
＋ (p11 + p21) log (p11 + p21)
＋ (p21 + p22) log (p21 + p22)
＋ (p12 + p22) log (p12 + p22) －2N log N
MLL_DM (P1, P2) ＝ p11 log p11 ＋ p12 log p12 ＋ p21 log p21 ＋ p22 log p22－N log N
However, N = p11 + p12 + p21 + p22
AIC_IM (P1, P2) = -2 x MLL_IM (P1, P2) + 2 x 2
AIC_IM (P1, P2): Subordinate AIC of a set of character attribute P1 and synonymous verb P2
AIC_DM (P1, P2) = -2 x MLL_DM (P1, P2) + 2 x 3
AIC_DM (P1, P2): Independent AIC of a pair of character attribute P1 and synonymous verb P2
V (P1, P2) = AIC_IM (P1, P2) -AIC_DM (P1, P2)
It is also preferable to make the computer function as such.

According to other embodiments in the program of the invention
The posted text is a text posted on an SNS (Social Networking Service) or a blog by an unspecified number of users.
It is also preferable to make the computer function so that the character attribute is a classification attribute based on any one of age, gender, birthplace, or a combination thereof.

According to other embodiments in the program of the invention
A means of dialogue in which an agent interacts with a user using a dialogue scenario,
Whether the utterance sentence replaced with the characteristic synonymous verb in the dialogue scenario causes a sense of discomfort to the user when or after the utterance sentence is spoken to the user during the dialogue with the user by the dialogue means. It is also possible to further function the computer as an utterance verb correction means that replaces the replaced feature synonymous verb with the original utterance verb when the user inquires by text or voice and determines that the user feels uncomfortable. preferable.

According to the dialogue scenario creating device of the present invention.
Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
Post-sentence storage means for accumulating multiple posts by multiple users for each character attribute,
Meaning element determination means for determining the meaning element in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs with the same meaning element.
Characteristic synonymous verb selection means for selecting a synonymous verb whose appearance frequency is most biased among multiple synonymous verbs in the same meaning element as a characteristic synonymous verb using a user's post based on the second character attribute. When,
It is characterized by having a verb replacement means for replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic synonymous verb to be the utterance sentence of the dialogue scenario of the second character attribute. And.

According to the dialogue scenario creation method of the present invention.
The device is
Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
It has a post sentence storage means for accumulating a plurality of post sentences by a plurality of users for each character attribute.
The first step of determining the meaning elements in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs having the same meaning element,
The second step of selecting the synonymous verb whose appearance frequency is most biased among multiple synonymous verbs of the same meaning as the characteristic synonymous verb by using the user's posted sentence based on the second character attribute.
The third step of replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic synonymous verb and making it the utterance sentence of the dialogue scenario of the second character attribute is executed. It is characterized by.

According to the program, device, and method of the present invention, it is possible to create a dialogue scenario in which verbs are replaced according to character attributes.

It is explanatory drawing which shows the uncomfortable feeling which a user has by the dialogue with an agent. It is a functional block diagram of the dialogue scenario creation apparatus in this invention. It is explanatory drawing of the dialogue scenario storage part and the post sentence storage part in this invention. It is a functional block diagram of the meaning element determination part in this invention. It is explanatory drawing of the characteristic type verb selection part in this invention. It is explanatory drawing of the verb substitution part in this invention. It is explanatory drawing which shows the dialogue scenario replaced by this invention. It is explanatory drawing which shows the interface which modifies the verb replaced by this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a functional configuration diagram of the dialogue scenario creating device in the present invention.

The dialogue scenario creation device 1 of the present invention automatically creates a dialogue scenario according to the character attribute of the agent.
According to FIG. 2, the dialogue scenario creating device 1 includes a dialogue unit 100, a dialogue scenario storage unit 101, a posted sentence storage unit 102, a meaning element determination unit 11, a characteristic synonymous verb selection unit 12, and a verb replacement. It has a unit 13 and a spoken verb correction unit 14. These functional components are realized by executing a program that makes a computer mounted on the device function. In addition, the processing flow of these functional components can be understood as a method for creating a dialogue scenario.

The dialogue scenario creating device 1 of the present invention is
The meaning element related to the verb included in the utterance sentence of the first dialogue scenario in the agent of the first character attribute is determined (meaning element determination unit 11).
From among the multiple synonymous verbs that exist in the meaning element, select the synonymous verb that characterizes the second character attribute of the second dialogue scenario that you want to create (characteristic synonymous verb selection unit 12).
The verb included in the utterance sentence of the first dialogue scenario is replaced with a synonymous verb corresponding to the second character attribute (verb replacement part 13).
By doing so, the second dialogue scenario in the agent of the second character attribute is automatically created.

[Dialogue unit 100]
The dialogue unit 100 activates an agent that interacts with the terminal 2 owned by the user by voice or text via the user interface. The user can select an agent according to a desired character attribute (age, gender, hometown, etc.). The agent refers to the dialogue scenario storage unit 101 and interacts with the user using the dialogue scenario according to the character attribute. The agent can be recognized by the user as a person based on the character attribute.

[Dialogue scenario storage unit 101]
The dialogue scenario storage unit 101 stores dialogue scenarios for each character attribute.
The dialogue scenario storage unit 101 constitutes a dialogue scenario with an received sentence and an uttered sentence. The utterance sentence assumes a text that the user will utter, and the utterance sentence is a text that the agent utters with respect to the utterance sentence.
Further, the character attribute of the agent is, for example, a classification attribute based on any one of age, gender, hometown, or a combination thereof.

FIG. 3 is an explanatory diagram of a dialogue scenario storage unit and a posted text storage unit in the present invention.

According to FIG. 3A, the dialogue scenario of the first character attribute stored in the dialogue scenario storage unit 101 is shown. Here, the first character attribute is, for example, a woman in her twenties.
(Received sentence) (Spoken sentence)
・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・
"What do you do with rice at night?""I'll eat fried rice."
・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・
The verb of the agent's utterance is "I'll eat", which is a way of speaking assuming the first character attribute "female in her twenties". According to the prior art, it was created manually in order to adjust the utterance sentence of a delicate verb according to the character attribute.
According to the embodiment of the present invention, the second dialogue scenario of the second character attribute "teens / male" is automatically created from the first dialogue scenario of the first character attribute "twenties / female".

[Posted text storage unit 102]
The posted text storage unit 102 stores a plurality of posted texts by a plurality of users for each character attribute (user attribute, demographic).
The posted text may be, for example, a text posted on an SNS (Social Networking Service) or a blog, and may be sent by an unspecified number of third parties or a specific person.

According to FIG. 3B, for example, the following posted sentences are listed for each character attribute.
Posts based on the first character attribute (female in their 20s)
"I'll eat fried rice"
・ ・ ・ ・ ・ ・
"I'm fine eating fried rice."
"Let's eat fried rice"
Posts based on the second character attribute (teens / male)
・ ・ ・ ・ ・ ・
"I'll eat"
・ ・ ・ ・ ・ ・
"Rice, eat and be full"
"Would you like to eat fried rice?"
Posts have a masculine tone, a feminine tone, a youthful tone, an elderly tone, and a local dialect-like tone. These affect the verbs in the post depending on the character attributes.

In order to classify an unspecified number of posted texts by character attribute, for example, as in Patent Document 1, a technique of estimating character attributes such as gender / age and birthplace can be used for each posted text. The profile (character attribute) of the poster is estimated using the exchange relationship between the posters that can be obtained from the SNS site server.
The character attribute of the posted text shall be the same classification attribute as the character attribute of the dialogue scenario. Character attributes affect verbs (eg, flexion) in particular in spoken sentences.

The posted text of the first character attribute is a posted text posted by a user (unspecified majority or specific person) having the same or similar character attribute as the first dialogue scenario to be replaced.
The post text of the second character attribute is a post text posted by a user (unspecified majority or specific person) with the same or similar character attribute as the second dialogue scenario to be replaced.

[Meaning element determination unit 11]
The meaning element determination unit 11 determines the "meaning element" in the "utterance verb" described in the utterance sentence of the dialogue scenario based on the first character attribute, and associates a plurality of "synonymous verbs" having the same meaning element. .. Here, when a plurality of meaning elements are determined for the "utterance verb" described in the utterance sentence of the dialogue scenario based on the first character attribute, one of the meaning elements is determined.

A "sememe" is a unit based on a concept that constitutes the meaning of a word. In linguistics, "meaning of sentence" and "meaning of word" are used properly. As an element that constitutes the "meaning" of a word, there is a meaning element according to a phoneme or a morpheme.

FIG. 4 is a functional configuration diagram of the meaning element determination unit in the present invention.

According to FIG. 4, the meaning element determination unit 11 inputs the dialogue scenario of the first character attribute, and outputs the meaning element table composed of the determined meaning elements and synonyms.
For example, it is assumed that a dialogue scenario of a female agent in her twenties is input as a dialogue scenario of the first character attribute to be input.
(Utterance)
"I'll eat fried rice"

According to FIG. 4, the meaning element determination unit 11 has a part of speech extraction unit 111, a synonymous verb search unit 112, and a co-occurrence frequency search unit 113.

<Part of speech extraction unit 111>
The part-of-speech extraction unit 111 extracts nouns and verbs that have a dependency relationship from the utterance sentences of the dialogue scenario based on the first character attribute.
The part-speech extraction unit 111 may use, for example, the Japanese dependency analyzer CaboCha. When a sentence is input to such an analyzer, nouns and verbs are output.
According to FIG. 4, the part of speech extraction unit 111 extracts nouns and verbs as follows.
(Utterance) (Noun) (Verb)
"I'll eat fried rice""Friedrice""Eat"
The part-speech extraction unit 111 outputs the verb "eat" to the synonymous verb search unit 112, and outputs the noun "fried rice".

<Synonymous verb search unit 112>
The synonym verb search unit 112 searches for a "synonym verb" similar to a dependency verb for each meaning element by using the "synonym corpus".

The synonym corpus is composed of meaning elements and synonyms, and includes, for example, the following.
・ Weblio Thesaurus (see Non-Patent Document 3)
-Associative thesaurus (see Non-Patent Document 4)
-Synonyms dictionary (see Non-Patent Document 5)
For example, when the verb "eat" is input, the synonymous verb search unit 112 searches for one or more synonymous verbs similar to the verb "eat".
At this time, the verb "eat" has the meaning (meaning) of "eating" such as "eating rice", "eating fish", and "eating vegetables", and "eating in the president's business" and "eating". There is a meaning of "self-supporting" such as "do what you like and eat". Then, a plurality of synonymous verbs are searched for each meaning element.
(Verb) (Meaning element) (Synonymous verb)
"Eat""Eat" Eat
Eat
To eat
To enjoy
"Self-supporting" to live
Live
Eat Multiple synonymous verbs "eat", "eat", "eat", and "eat" have the same meaning "eat". In addition, the multiple synonymous verbs "living", "living", and "eating" have the same meaning "self-supporting".

The synonymous verb search unit 112 derives all meaning elements for the input verb, and searches for all synonymous verbs for each meaning element. The searched meaning element table is output to the co-occurrence frequency search unit 113.

<Co-occurrence frequency search unit 113>
The co-occurrence frequency search unit 113 searches for "co-occurrence frequency between the dependent noun and each synonymous verb" from the user's posted sentences based on the first character attribute, and searches for the co-occurrence frequency of each synonymous verb for each meaning element. Count the frequency of appearance. Then, the meaning element with the highest co-occurrence frequency of synonymous verbs is determined.

According to FIG. 4, the co-occurrence frequency search unit 113 inputs, for example, the meaning element table of the verb “eat” from the synonymous verb search unit 112, and inputs the noun “fried rice” from the part of speech extraction unit 111. Further, the co-occurrence frequency search unit 113 refers to the posted sentence storage unit 102 and acquires the posted sentence of the first character attribute.
Here, the co-occurrence frequency search unit 113 determines a meaning element having a high frequency of co-occurrence between the noun "fried rice" and each synonymous verb in the meaning element table.

According to FIG. 4, it is assumed that the posted sentence storage unit 102 has eight posted sentences having the first character attribute. Among them, the noun "fried rice" and the synonymous verb synonymous with the meaning element "eat" co-occur in the following three posts.
(Posted text) (Noun) (Synonymous verb)
"Eat fried rice""Friedrice""Eat"
"Eat fried rice and feel fine""Friedrice""Eat"
"Let's eat fried rice""Friedrice""Eat"
According to FIG. 4, there are “eat” and “eat” as synonymous verbs that frequently co-occur with the noun “fried rice”. These synonymous verbs correspond to the meaning "eat". Then, when comparing the meaning elements "eat" and "self-support", the meaning element with the highest co-occurrence frequency is "eat".
That is, the co-occurrence frequency search unit 113 determines the meaning element "eat" and outputs a meaning element table associated with the synonymous verbs "eat", "eat", "eat", and "eat".

[Characteristic verb selection unit 12]
The feature synonymous verb selection unit 12 uses a user's post based on the second character attribute to select a synonymous verb whose appearance frequency is most biased among a plurality of synonymous verbs in the same meaning element. Select as.

The second character attribute is based on the agent of the newly created second dialogue scenario. The agent may remind the user of a specific person (eg, entertainer or personality). In that case, as the user's post based on the second character attribute, the post posted by the specific person in the past can also be used. This makes it possible to automatically create a second dialogue scenario based on a specific person from a general first dialogue scenario based on an unspecified person.

FIG. 5 is an explanatory diagram of a characteristic synonymous verb selection unit in the present invention.

According to FIG. 5, the meaning element table is input from the meaning element determination unit 11, and the posted text of the second character attribute based on the dialogue scenario to be created is referred to.
Here, it is assumed that a teenager / male is set as the second character attribute and a dialogue scenario of the agent is created.
The feature synonymous verb selection unit 12 searches for the frequency of appearance of the following plurality of synonymous verbs in the same meaning by using the user's posted sentence based on the second character attribute.
Meaning "Eat"
Synonymous verbs "eat""eat""eat""eat"
Then, the characteristic synonymous verb selection unit 12 searches for the post sentence of the second character attribute in which the synonymous verb is described for each synonymous verb, and selects the synonymous verb whose appearance frequency is most biased. Select as.

The feature synonymous verb selection unit 12 calculates the evaluation value V (P1, P2) of the synonymous verb P2 for the character attribute P1 using the Akaike's Information Criterion (AIC) as follows. The synonymous verb P2 whose evaluation value V (P1, P2) is equal to or greater than a predetermined threshold, or whose evaluation value V (P1, P2) is a predetermined number in descending order is targeted for detection.
p11: The number of posts related to the synonymous verb P2 by the user of the character attribute P1.
p12: The number of posted sentences other than the synonymous verb P2 by the user of the character attribute P1.
p21: The number of posts related to the synonymous verb P2 by users other than the character attribute P1.
p22: The number of posted sentences other than the synonymous verb P2 by users other than the character attribute P1.
p11 ＋ p12 ＋ p21 ＋ p22 ＝ Number of posts posted by all users
MLL_IM (P1, P2) = (p11 + p12) log (p11 + p12)
＋ (p11 + p21) log (p11 + p21)
＋ (p21 + p22) log (p21 + p22)
＋ (p12 + p22) log (p12 + p22) －2N log N
MLL_DM (P1, P2) ＝ p11 log p11 ＋ p12 log p12 ＋ p21 log p21 ＋ p22 log p22－N log N
However, N = p11 + p12 + p21 + p22
AIC_IM (P1, P2) = -2 x MLL_IM (P1, P2) + 2 x 2
AIC_IM (P1, P2): Subordinate AIC of a set of character attribute P1 and synonymous verb P2
AIC_DM (P1, P2) = -2 x MLL_DM (P1, P2) + 2 x 3
AIC_DM (P1, P2): Independent AIC of a pair of character attribute P1 and synonymous verb P2
V (P1, P2) = AIC_IM (P1, P2) -AIC_DM (P1, P2)

The feature synonymous verb selection unit 12 selects one synonymous verb according to the Akaike information criterion. Here, the top several synonymous verbs may be selected, and one synonymous verb may be randomly selected from them.

According to FIG. 5, when the Akaike information criterion for each synonymous verb is calculated for teenagers and males with character attributes, it can be understood that the frequency of appearance is biased toward the synonymous verb "eating".
Finally, the feature synonymous verb selection unit 12 outputs the verb “eat” and the synonymous verb “eat” to the verb replacement unit 13.

[Verb substitution part 13]
The verb replacement unit 13 replaces the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with the characteristic synonymous verb to obtain the utterance sentence of the dialogue scenario of the second character attribute.

FIG. 6 is an explanatory diagram of a verb substitution part in the present invention.

According to FIG. 6, the verb replacement unit 13 replaces the verb “eat” with the synonymous verb “eat” in the dialogue scenario of the first character attribute (female in her twenties), and the second character attribute (10). Create as a dialogue scenario for teens and men).
Speech text of the dialogue text of the first character attribute: "I will eat fried rice"
The utterance of the dialogue text of the second character attribute: "I will eat fried rice"

As described above, by repeating the meaning element determination unit 11, the feature synonymous verb selection unit 12, and the verb replacement unit 13, finally, from the dialogue scenario of different character attributes, the dialogue scenario of the desired character attribute Can be created automatically.

FIG. 7 is an explanatory diagram showing a dialogue scenario replaced by the present invention.

According to FIG. 7, a dialogue scenario with a character attribute of a teenager / male is created from a dialogue scenario with a character attribute of a woman in her twenties as shown in FIG. 1 (a).
User: "What do you do with rice at night?"
Agent: "I'll eat fried rice."
The user is interacting with an agent with a teenage / male character attribute, and does not feel uncomfortable when he / she is told "I will eat". This is because teenagers and men often make rough utterances.

Here, from the user's point of view, Japanese may have an unnatural sense of discomfort with respect to the utterance sentence of the dialogue scenario in which the verb is replaced. For example, suppose that the utterance sentence of the dialogue scenario is as follows.
"I want to eat fried rice"-> Verb: "Eat", Function word: "Tai"
When the synonymous verb "eat" to be replaced is selected, the following utterance sentence is created.
"I want to eat fried rice"->"I want to eat fried rice"
In this way, because the verb is replaced without considering the inflected form of the verb, a sense of incongruity arises in Japanese.

FIG. 8 is an explanatory diagram showing an interface for modifying a verb replaced by the present invention.

[Utterance verb correction part 14]
The utterance verb correction unit 14 tells the user when or after the utterance sentence replaced with the characteristic synonymous verb in the dialogue scenario is spoken to the user during the dialogue with the user in the dialogue unit 100. Inquire by text or voice whether or not the utterance feels strange. At this time, the utterance verb correction unit 14 transmits inquiry information to the user interface of the terminal 2 via the dialogue unit 100.
According to FIG. 8, the following two-choice inquiry button is displayed on the interactive application of the terminal 2.
"There is something wrong with the reply"
"No discomfort in reply"

When the user determines that a feeling of strangeness is generated, "there is a feeling of strangeness in the reply" is tapped. This tap information is transmitted from the terminal 2 to the dialogue scenario creating device 1. The tap information is recognized by the utterance verb correction unit 14, and the replaced feature synonymous verb is replaced again to return to the original utterance verb.
In this way, the agent can interact with the user to determine the validity of the verb substitution.

As described above in detail, according to the program, device, and method of the present invention, it is possible to create a dialogue scenario in which verbs are replaced according to character attributes. This makes it possible to automatically create a dialogue scenario that does not feel strange as a character attribute.

Various modifications, modifications and omissions of the above-mentioned various embodiments of the present invention within the scope of the technical idea and viewpoint of the present invention can be easily carried out by those skilled in the art. The above explanation is just an example and does not attempt to limit anything. The present invention is limited only to the scope of claims and their equivalents.

1 Dialogue scenario creation device 100 Dialogue section 101 Dialogue scenario storage section 102 Post sentence storage section 11 Significance element determination section 111 Part of speech extraction section 112 Synonymous verb search section 113 Co-occurrence frequency search section 12 Characteristic verb substitution section 14 Speech verb correction part 2 Terminal

Claims (8)

Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
Post-sentence storage means for accumulating multiple posts by multiple users for each character attribute,
Meaning element determination means for determining the meaning element in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs with the same meaning element.
Characteristic synonymous verb selection means for selecting a synonymous verb whose appearance frequency is most biased among multiple synonymous verbs in the same meaning element as a characteristic synonymous verb using a user's post based on the second character attribute. When,
Replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic verb, and making the computer function as a verb replacement means to be the utterance sentence of the dialogue scenario of the second character attribute. A program featuring. Meaning factor determination means
When multiple meanings are determined for the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute,
Using the synonym corpus, search for multiple synonymous verbs that have the same meaning for each meaning element.
In the user's post based on the first character attribute, the frequency of appearance of each synonymous verb is counted for each meaning element, and
The program according to claim 1, wherein the computer functions so as to determine the meaning element having the highest frequency of appearance of each synonymous verb. Meaning factor determination means
Part-of-speech extraction means for extracting nouns and verbs that are dependent on each other from the utterance sentences of the dialogue scenario based on the first character attribute.
A synonym verb search method for searching for synonymous verbs similar to dependent verbs for each meaning element using a synonym corpus,
The co-occurrence frequency of searching the co-occurrence frequency of the dependent noun and each synonymous verb from the user's post based on the first character attribute, and determining the co-occurrence element with the highest co-occurrence frequency of each synonymous verb. The program according to claim 1 or 2, wherein the computer functions as a search means. Characteristic verb selection means
As shown below, using the Akaike's Information Criterion (AIC), the evaluation value V (P1, P2) of the synonymous verb P2 for the character attribute P1 is calculated, and the evaluation value V (P1, P2). The synonymous verb P2 whose evaluation value V (P1, P2) is greater than or equal to the predetermined threshold or whose evaluation value V (P1, P2) is the largest is the predetermined number.
p11: The number of posts related to the synonymous verb P2 by the user of the character attribute P1.
p12: The number of posted sentences other than the synonymous verb P2 by the user of the character attribute P1.
p21: The number of posts related to the synonymous verb P2 by users other than the character attribute P1.
p22: The number of posted sentences other than the synonymous verb P2 by users other than the character attribute P1.
p11 ＋ p12 ＋ p21 ＋ p22 ＝ Number of posts posted by all users
MLL_IM (P1, P2) = (p11 + p12) log (p11 + p12)
＋ (p11 + p21) log (p11 + p21)
＋ (p21 + p22) log (p21 + p22)
＋ (p12 + p22) log (p12 + p22) －2N log N
MLL_DM (P1, P2) ＝ p11 log p11 ＋ p12 log p12 ＋ p21 log p21 ＋ p22 log p22－N log N
However, N = p11 + p12 + p21 + p22
AIC_IM (P1, P2) = -2 x MLL_IM (P1, P2) + 2 x 2
AIC_IM (P1, P2): Subordinate AIC of a set of character attribute P1 and synonymous verb P2
AIC_DM (P1, P2) = -2 x MLL_DM (P1, P2) + 2 x 3
AIC_DM (P1, P2): Independent AIC of a pair of character attribute P1 and synonymous verb P2
V (P1, P2) = AIC_IM (P1, P2) -AIC_DM (P1, P2)
The program according to any one of claims 1 to 3, wherein the computer functions as described above. The posted text is a text posted on an SNS (Social Networking Service) or a blog by an unspecified number of users.
The program according to any one of claims 1 to 4, wherein the character attribute causes the computer to function as a classification attribute based on any one of age, gender, and birthplace, or a combination thereof. A means of dialogue in which an agent interacts with a user using a dialogue scenario,
Whether the utterance sentence replaced with the characteristic synonymous verb in the dialogue scenario is uttered to the user during the dialogue with the user by the dialogue means, or after that, the utterance sentence causes discomfort to the user. When the user inquires by text or voice and determines that the user feels uncomfortable, the computer can be further functioned as an utterance verb correction means for replacing the replaced feature synonymous verb with the original utterance verb. The program according to any one of claims 1 to 5, which is characterized. Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
Post-sentence storage means for accumulating multiple posts by multiple users for each character attribute,
Meaning element determination means for determining the meaning element in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs with the same meaning element.
Characteristic synonymous verb selection means for selecting a synonymous verb whose appearance frequency is most biased among multiple synonymous verbs in the same meaning element as a characteristic synonymous verb using a user's post based on the second character attribute. When,
It is characterized by having a verb replacement means for replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic synonymous verb to be the utterance sentence of the dialogue scenario of the second character attribute. Dialogue scenario creation device. It is a method of creating a dialogue scenario of the device.
The device is
Dialogue scenario storage means that accumulates dialogue scenarios for each character attribute,
It has a post sentence storage means for accumulating a plurality of post sentences by a plurality of users for each character attribute.
The first step of determining the meaning elements in the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute, and associating a plurality of synonymous verbs having the same meaning element,
The second step of selecting the synonymous verb whose appearance frequency is most biased among multiple synonymous verbs of the same meaning as the characteristic synonymous verb by using the user's posted sentence based on the second character attribute.
The third step of replacing the utterance verb described in the utterance sentence of the dialogue scenario based on the first character attribute with a characteristic synonymous verb and making it the utterance sentence of the dialogue scenario of the second character attribute is executed. How to create a dialogue scenario featuring.

Images