JP2004258902A - Conversation controller, and conversation control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To preliminarily output a reply sentence related to a specific subject of conversation desired to be developed, so as to guide the present subject of conversation to the specific subject of conversation, when a user's speech content is recognized with a plurality of meanings. <P>SOLUTION: This controller/method is provided with a conversation subject retrieving part 360 for retrieving each phrase containing the speech content out of the plurality of phrases preliminarily stored, based on the speech content input from the user, a selection part 380 for selecting preliminarily the phrase related to the subject of conversation desired to be developed out of the respective phrases, based on the each phrase retrieved by the subject retrieval part 360, and a reply sentence retrieval part 370 for retrieving the reply sentence corresponding to the phrase selected by the selection part 380. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a conversation control device and a conversation control method.
[0002]
[Prior art]
A conventional conversation control device can output a pre-stored answer sentence corresponding to the utterance content based on the utterance content from the user (for example, Patent Document 1).
[0003]
[Patent Document 1]
JP-A-2002-169804 (pages 6 to 15, FIG. 16)
[0004]
[Problems to be solved by the invention]
However, when the content of the user's utterance is captured in a plurality of meanings and there are a plurality of answer sentences corresponding to the utterance content, the conversation control device generally outputs the answer sentence searched first. Was. For this reason, when the utterance content of the user is captured in a plurality of meanings, the conversation control device selects, from the respective meanings, the meaning of the utterance content related to a specific topic to be developed in advance, and selects the meaning. An answer sentence corresponding to the meaning of the utterance content could not be output.
[0005]
For example, when the utterance content from the user is "Please tell me IR", the utterance content is captured in the meaning of IR (Inazuma Rush) of the effect image on the gaming machine or IR (Investor Relations) of the company. Sometimes. In this case, when the conversation control device first searches for an answer sentence corresponding to the IR (Inazuma Rush) of the effect image on the gaming machine from among the answer sentences, it outputs the first searched answer sentence. .
[0006]
That is, even when the conversation control device wants to output an answer sentence corresponding to the IR of the company, when the answer sentence corresponding to the IR (Inazuma Rush) of the effect image is first searched, the first searched answer The statement was output automatically. For this reason, there has been a demand for the development of a system capable of outputting an answer sentence related to a specific topic to be developed in advance when the utterance content from the user is captured in a plurality of meanings.
[0007]
Therefore, the present invention has been made in view of the above points, and when the utterance content of the user is captured in a plurality of meanings, by outputting an answer sentence related to a specific topic to be developed in advance, It is an object to provide a conversation control device and a conversation control method that can guide a current topic to a specific topic.
[0008]
[Means for Solving the Problems]
The present invention has been made in order to solve the above-described problem, and when searching for an answer sentence corresponding to input information based on input information including utterance content input from the user, the present invention has been made to solve the problem. Searches each phrase including the input information from a plurality of phrases stored in advance based on the input information, and selects a phrase related to a topic to be developed in advance from each phrase based on the searched phrases. Then, an answer sentence corresponding to the selected phrase is searched.
[0009]
That is, in the present invention, the phrase includes second morpheme information indicating a morpheme composed of one character, a plurality of character strings, or a combination thereof, and the plurality of second morpheme information includes an answer sentence to the user. And a priority order to be selected as an answer sentence are associated with each other, each second morpheme information and each answer sentence are stored in advance, and based on a character string included in the input information input by the user, a character Extracting at least one morpheme constituting the minimum unit of the column as first morpheme information, based on the extracted first morpheme information, from among the previously stored second morpheme information, including the first morpheme information When the second morpheme information is searched, and when a plurality of second morpheme information is searched, each of the searched second morpheme information is determined in accordance with the magnitude of each priority order associated with each searched second morpheme information. Dimorphological information Elect one second morpheme information from, based on the second morpheme information elected, characterized by searching an answer sentence associated with the second morpheme information.
[0010]
According to the invention of the present application, when the second morpheme information 1 and the second morpheme information 2 are searched as the second morpheme information including the extracted first morpheme information, the conversation control device: From the retrieved second morpheme information 1 and second morpheme information 2, for example, the second morpheme information 1 associated with a higher priority is selected, and the answer sentence associated with the selected second morpheme information 1 is selected. Can be output. Here, the fact that a plurality of phrases each including input information is searched, or that a plurality of second morphological information including first morphological information constituting the input information are searched for, means that the input information has a plurality of meanings. Means that.
[0011]
That is, when the input information is captured in a plurality of meanings, and when the plurality of second morpheme information (or phrases) is searched, the conversation control device expands the second morpheme information from the respective second morpheme information in advance by the conversation control device. It is possible to select second morpheme information (second morpheme information associated with a high priority) related to a topic to be performed.
[0012]
As a result, the conversation control device does not output a passive answer sentence based on the content of the utterance from the user (for example, faithfully answer the content of the question from the user). It is possible to output an active answer sentence that expands to a specific topic, in other words, to output an answer sentence that guides the current topic to the specific topic.
[0013]
In the above invention, if each of the second morpheme information searched this time is searched for more than a predetermined reference number, the second morpheme information different from the previously selected second morpheme information is searched this time. May be selected from the second morpheme information. Thereby, if the second morpheme information selected last time is included in each of the second morpheme information searched this time, and each of the second morpheme information searched this time is searched beyond a predetermined reference number, the conversation The control device can select the second morpheme information different from the second morpheme information selected last time from each of the second morpheme information searched this time.
[0014]
That is, when the same utterance content is repeatedly uttered by the user, the conversation control device can sequentially change the second morpheme information according to the number of times of the repetition. As a result, the conversation control device can change the answer sentence for guiding the current topic to the specific topic one by one according to the number of repeated utterance contents.
[0015]
In the above invention, the answer sentence may include a content for guiding the user to a specific topic. In this case, since the content to guide the user to a specific topic is included in the answer sentence, the conversation control device outputs the answer sentence so that the current topic can be more accurately adjusted to the topic to be developed in advance. Can be guided. In other words, the conversation control device can take the initiative on the current topic internally and guide the current topic more accurately to a specific story.
[0016]
In the above invention, when the second morpheme information searched last time is included in each second morpheme information searched this time, the second morpheme information searched last time is selected from the second morpheme information searched this time. It is preferable to specify a priority next to the priority associated with the information, and select second morpheme information associated with the specified priority.
[0017]
BEST MODE FOR CARRYING OUT THE INVENTION
[First embodiment]
(Basic configuration of conversation control system)
A conversation control system according to the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of a conversation control system having a conversation control device 1 according to the present embodiment.
[0018]
As shown in FIG. 1, the conversation control device 1 includes an input unit 100, a speech recognition unit 200, a conversation control unit 300, a sentence analysis unit 400, a conversation database 500, an output unit 600, and a speech recognition dictionary storage. And a unit 700.
[0019]
In the present embodiment, for convenience of explanation, the description will be limited to the utterance content of the user (this utterance content is a type of input information). However, the present invention is not limited to the utterance content of the user, and is not limited to the utterance content of the user. For example, the input information may be input from a user or the like. Therefore, the "speech content" described below can also be described by replacing "speech content" with "input information".
[0020]
Similarly, in the following description, for convenience of explanation, the description will be limited to “speech sentence type” (speech type). The “input type” indicating the type of input information that has been input may be used. Therefore, the "utterance sentence type" (utterance type) described below can also be described by replacing "utterance type" with "input type".
[0021]
The input unit 100 is an obtaining unit that obtains input information from a user. In the present embodiment, the input unit 100 includes a microphone, a keyboard, and the like. The input unit 100 is also a character recognition unit that specifies a character string indicating input information based on input information input by a user.
[0022]
Here, the input information means characters, symbols, voices and the like input through a keyboard or the like. Specifically, the input unit 100 acquires input information (other than voice) of the user and outputs the acquired input information to the conversation control unit 300. Also, the input unit 100 that has acquired the utterance content from the user (this utterance content is composed of voice and is a type of input information) with a microphone or the like, the input unit 100 converts the voice constituting the acquired utterance content into an audio signal. Is output to the voice recognition unit 200.
[0023]
The voice recognition unit 200 is a character recognition unit that specifies a character string included in the utterance content based on the utterance content acquired by the input unit 100. Specifically, the voice recognition unit 200 to which the voice signal is input from the input unit 100 analyzes the input voice signal, and stores a character string included in the analyzed voice signal in the voice recognition dictionary storage unit 700. The specified character string is output to the conversation control unit 300 as a character string signal. The voice recognition dictionary storage unit 700 stores a dictionary corresponding to a standard voice signal.
[0024]
The sentence analysis unit 400 analyzes a character string specified by the input unit 100 or the speech recognition unit 200. In the present embodiment, as shown in FIG. 2, a morpheme extraction unit 410 and a phrase analysis unit 420 , A sentence structure analysis unit 430, an utterance type determination unit 440, a morpheme database 450, and an utterance type database 460.
[0025]
The morpheme extraction unit 410 is a morpheme extraction unit that extracts, as first morpheme information, at least one morpheme constituting the minimum unit of the character string based on the character string specified by the input unit 100 or the speech recognition unit 200.
[0026]
Specifically, the morpheme extraction unit 410 to which the character string signal is input from the management unit 310 extracts each morpheme from the character string corresponding to the input character string signal. Here, in the present embodiment, a morpheme means a minimum unit of a word configuration represented in a character string. As shown in FIG. 3, examples of the minimum unit of the word configuration include a part of speech such as a noun, an adjective, and a verb. In the present embodiment, each morpheme is represented as m1, m2,..., Ml.
[0027]
That is, the morpheme extraction unit 410 collates a character string corresponding to the input character string signal with a morpheme group such as a noun, adjective, or verb stored in the morpheme database 450 in advance, and extracts a morpheme from the character string. Each morpheme (m1, m2,...) That matches the group is extracted, and each extracted morpheme is output to the phrase analysis unit 420 as an extraction signal.
[0028]
The phrase analysis unit 420 is a conversion unit that converts each morpheme into a phrase format based on each morpheme extracted by the morpheme extraction unit 410. Specifically, the phrase analysis unit 420 to which the extraction signal is input from the morpheme extraction unit 410 uses the respective morphemes corresponding to the input extraction signal to combine them into a phrase format.
[0029]
In the present embodiment, the phrase form refers to a sentence having an independent word or an independent word with one or more attached words in the Japanese grammar, or a character string that does not break the meaning of the Japanese grammar. It means a sentence that is as small as possible. This clause is expressed as p1, p2,... Pk in the present embodiment.
[0030]
That is, as shown in FIG. 4, the phrase analyzing unit 420 converts the dependency element of each morpheme (for example, (m2) · (m4) · (m5) based on each morpheme corresponding to the input extracted signal. ) ..) are extracted, and each morpheme is grouped into each phrase based on the extracted dependency elements. "T" shown in the figure means transposition.
[0031]
The phrase analysis unit 420 that compiles each morpheme into each phrase, the sentence structure analysis unit 430 and the utterance type determination unit 440, using the morphological information including each morpheme that summarizes each morpheme and each morpheme that composes each phras as a morphological signal. Output to
[0032]
The sentence structure analysis unit 430 is a classification unit that classifies each morpheme of the first morpheme information segmented by the phrase analysis unit 420 into attributes such as subject matter and object case. Specifically, the sentence structure analysis unit 430 to which the sentence pattern signal is input from the sentence pattern analysis unit 420, based on each morpheme corresponding to the input sentence pattern signal and the sentence formed by each morpheme, Determine the "case composition".
[0033]
Here, the “case composition” means a case (attribute) indicating a substantial concept in a phrase, and in the present embodiment, for example, a subject (subject case) meaning a subject / subject and an object Examples include a meaningful object (target case), an action meaning an action, a time meaning a time (tense, aspect), a location meaning a place, and the like. In the present embodiment, each morpheme associated with three elements “case” (case configuration) of a subject, an object, and an action is defined as first morpheme information.
[0034]
That is, as shown in FIG. 5, for example, when the dependency element of each morpheme is "" or "ha", the sentence structure analysis unit 430 determines that the morpheme before the dependency element is the subject (subject) Or nominative). In addition, for example, when the dependency element of each morpheme is "" or "", the sentence structure analysis unit 430 determines that the morpheme before the dependency element is an object (object).
[0035]
Further, for example, when the dependency element of each morpheme is “do”, the sentence structure analysis unit 430 determines that the morpheme preceding the dependency element is an action (predicate; this predicate is composed of a verb, an adjective, etc.). Is determined.
[0036]
The sentence structure analysis unit 430 that has determined the “case configuration” of each morpheme constituting each phrase, based on the first morpheme information associated with the determined “case configuration”, determines the range of a topic described below. A topic search command signal for specifying is output to the reflective determination unit 320.
[0037]
The utterance type determination unit 440 is a type specifying unit that specifies an utterance type (input type) indicating the type of utterance content (input information) based on the phrase specified by the phrase analysis unit 420. Specifically, the “speech sentence type” (speech type) is determined based on each morpheme corresponding to the sentence pattern signal input from the phrase analysis unit 420 and the clause composed of each morpheme.
[0038]
In this embodiment, as shown in FIG. 6, the “utterance sentence type” is a statement (D; Declaration), a comment sentence (I; Impression), a conditional sentence (C; Condition), and a result sentence (D: Declaration). E; Effect), time sentence (T; Time), place sentence (L; Location), repulsion sentence (N; Negation), and the like. As shown in FIG. 6, each of these sentences is expressed as an affirmative sentence (A; Answer) or a question sentence (Q; Question).
[0039]
The statement is a sentence composed of a user's opinion or idea, and in the present embodiment, as shown in FIG. 6, a sentence such as "I like Sato" is exemplified. The impression sentence means a sentence composed of impressions held by the user. The location sentence means a sentence composed of locational elements.
[0040]
The result sentence means a sentence composed of a sentence including a result element for a topic. The time sentence means a sentence including a temporal element related to a topic.
[0041]
The conditional sentence means a sentence including elements such as a premise of the topic and a condition or a reason why the topic is established when one utterance is regarded as a topic. The repulsive sentence means a sentence including an element that repels the utterer. An example sentence for each “utterance sentence type” is as shown in FIG.
[0042]
That is, based on each phrase corresponding to the input sentence pattern signal, the utterance type determination unit 440 collates each phrase with each dictionary stored in the utterance type database 460, and from among each phrase, Extract sentence elements related to the dictionary. The utterance type determination unit 440 that extracts a sentence element related to each dictionary from each phrase determines a “speech sentence type” based on the extracted sentence element. The sentence element means a minute type for specifying the type of the character string, and in the present embodiment, the sentence element includes the above-described definition phrase (to).
[0043]
Here, the utterance type database 460 includes, as shown in FIG. 7, a definition expression example dictionary including a dictionary related to a definition phrase (for example, 〜), and a positive phrase (for example, agree, agree, ping-pong). positive cases dictionary with a dictionary related to, a result clause (for example, so, so) the results representation case dictionary with a dictionary related to, greeting clause (for example, Hello) greeting case dictionary with a dictionary related to, It is composed of a negative case dictionary having a dictionary related to a negative phrase (for example, it is not stupid, opposite), and each dictionary is associated with a “speech sentence type”.
[0044]
Accordingly, the utterance type determination unit 440 compares the phrase with each dictionary stored in the utterance type database 460, extracts a sentence element related to each dictionary from the phrase, and associates the extracted sentence element with the extracted sentence element. The “utterance sentence type” can be determined by referring to the determined determination type.
[0045]
The utterance type determination unit 440 outputs an answer search command signal for causing the corresponding user to search for a specific answer sentence to the answer sentence search unit 370 based on an instruction from the topic search unit 360 described later.
[0046]
The conversation database 500 stores a plurality of pieces of second morpheme information indicating each morpheme composed of one character, a plurality of character strings, or a combination thereof, and a plurality of answer sentences to the user with respect to the utterance content in advance. Storage means (discussion storage means; morphological storage means). The conversation database 500 is also an answer storage unit (discussion storage unit) that stores a plurality of answer types indicating the types of each answer sentence associated with a plurality of answer sentences in advance in association with the second morpheme information.
[0047]
Further, the conversation database 500 previously stores a plurality of discourse ranges (keywords) indicating morphemes constituting ranges related to the input contents to be input from the user or the answer sentence to the user. But also. The discourse range (keyword) is associated with a plurality of pieces of second morpheme information indicating morphemes composed of one character, a plurality of character strings, or a combination thereof. Each second morpheme information includes an answer sentence to the user. Associated with each.
[0048]
Furthermore, the conversation database 500 is also an answer storage unit (discussion storage unit) that classifies and stores each element constituting the second morphological information into attributes such as a subject composed of a nominative case and a target case composed of a purpose case. .
[0049]
As shown in FIG. 8, in the present embodiment, the conversation database 500 is roughly divided into a discourse meaning a range that is relevant to the utterance contents that will be uttered by the user or the answer sentence to the user. It is composed of a range (discourse) and a topic (topic) indicating a range most closely related to the content spoken by the user. As shown in the drawing, the “discussion range” is positioned as a higher concept of “topic” in the present embodiment.
[0050]
Each discourse range can be configured to have a hierarchical structure, as shown in FIG. As shown in the figure, for example, the discourse range (entertainment) of a higher concept with respect to a certain discourse range (movie) is arranged in an upper hierarchical structure, and the discourse range of a lower concept with respect to the discourse range (movie) (movie) Attribute, showing movie) can be located in a lower hierarchical structure. That is, in the present embodiment, each discourse range can be arranged at a hierarchical position where the relationship between the superordinate concept, the subordinate concept, the synonym, and the synonym with the other discourse ranges is clear.
[0051]
As described above, the discourse range is composed of each topic. In the present embodiment, for example, if the discourse range is the A movie name, the discourse range includes a plurality of topics related to the A movie name.
[0052]
This topic means a morpheme composed of one character, a plurality of character strings, or a combination thereof, that is, each morpheme constituting the utterance content that will be uttered by the user. In the present embodiment, It is composed of each morpheme associated with the subject (subject), the object (target case), and the "case" (attribute) of the action. In the present embodiment, each morpheme associated with these three elements is expressed as a topic title (this topic title corresponds to a lower concept of “topic”) (second morpheme information). .
[0053]
The topic title is not limited to each morpheme associated with the above three elements, but may include other "cases", that is, time (tense, aspect) meaning time, location meaning location, Each morpheme may be associated with a condition indicating a condition, an impression indicating an impression, an effect indicating a result, or the like.
[0054]
In this embodiment, the topic title (second morpheme information) is stored in the conversation database 500 in advance, and the first morpheme information (derived from the utterance content uttered by the user) They are distinct.
[0055]
For example, if the topic title has a discourse range of “A movie name”, as shown in FIG. 10, subject (A movie name), object (director), action (great) ｛this is “A movie The name director is made up of｝, which means “great.”
[0056]
If there is no morpheme associated with “case composition” (subject, object, action, etc.) in the topic title, this portion is indicated with “*” in the present embodiment.
[0057]
For example, what is the $ A movie name? When the sentence of｝ is converted into a topic title (subject; object; action), what is the name of ｛A movie? In the sentence of｝, “A movie name” can be specified as a subject, but other “objects” and “actions” are not elements of the sentence, so the topic title is “subject” (A movie name) No "object"(*); no "action" (*) (see FIG. 10).
[0058]
In this embodiment, the answer sentence is associated with each topic title (second morpheme information) (see FIG. 8). In this embodiment, as shown in FIG. 11, the answer sentence includes a statement (D; Declaration) and an impression sentence in order to make an answer corresponding to the type of the utterance sentence uttered by the user. (I; Impression), conditional sentence (C; Condition), result sentence (E; Effect), time sentence (T; Time), location sentence (L; Location), negative sentence (N; Negation), etc. (answer) Types).
[0059]
That is, as shown in FIG. 12, each answer sentence is associated with, for example, the discourse range (Sato) {lower concept; home run, higher concept; grass baseball, synonyms; panda Sato / Sato / Panda} and each topic title Have been.
[0060]
As shown in the figure, for example, the topic title 1-1 is $ (Sato; *; I like): This is the order of (subject; object; action) as described above. If the order is the same as in the following, the answer sentence 1-1 corresponding to the topic title 1-1 is (DA; statement positive sentence "I like Sato"), (IA; impression positive sentence) "I like Sato very much"), (CA; positive statement on condition "Because Sato's home run is very impressive"), (EA; positive statement on result "I always watch TV games where Sato appears"), (TA; time affirmative sentence "I actually like it because of the five consecutive at-bats in Koshien"), (LA; place affirmative sentence "I like the serious face when standing on a blow"), (NA Affirmative sentence "I don't want to talk to someone who dislikes Sato, goodbye").
[0061]
In the present embodiment, as shown in FIG. 2, the conversation control unit 300 includes a management unit 310, a reflective determination unit 320, a parrot return determination unit 330, a discourse range determination unit 340, and an abbreviation sentence complementing unit 350. , A topic search unit 360 and an answer sentence search unit 370.
[0062]
The management section 310 controls the entire conversation control section 300. Specifically, the management unit 310 to which the character string has been input from the input unit 100 or the voice recognition unit 200 outputs the input character string to the morpheme extraction unit 410 as a character string signal. Further, management section 310 outputs the answer sentence obtained by answer sentence search section 370 to output section 600.
[0063]
The reflexive determination unit 320 is a fixed form acquisition unit that compares the first morpheme information extracted by the morpheme extraction unit 410 with each fixed content, and searches for fixed content including the first morpheme information from each fixed content. .
[0064]
Here, the standard content means reflective element information for answering the standard content to the utterance content from the user, and this reflective element information is stored in the reflective element database 801 (standard type storage means) in advance. A plurality is stored. As the reflective element information, in the present embodiment, as shown in FIG. 13, for example, "Good morning", "Hello", "Good evening", "greeting elements" such as "Hey", "I see", "Really?" And "typical elements".
[0065]
Specifically, the reflexive determination unit 320 to which the topic search command signal has been input from the sentence structure analysis unit 430 receives the first morphological information included in the input topic search command signal and each of the information stored in the reflection element database 801. The reflection element information is collated with the reflection element information, and the reflection element information including the first morphological information is retrieved from the respective reflection element information, and the retrieved reflection element information is output to the management unit 310.
[0066]
That is, assuming that the reflection element information is D1 and the first morpheme information is W, the reflection determining unit 320 determines that the relation of W∩D1 ≠ φ (φ; empty set) is satisfied, Perform a process to make a typical answer.
[0067]
For example, when the user utters the utterance content of “good morning”, the reflexive determination unit 320 compares the utterance content “good morning” with each piece of reflection element information, and from each piece of reflection element information, The reflection element information “good morning” including (matching with) “good morning” is retrieved, and the retrieved reflection element information “good morning” is output to the management unit 310.
[0068]
When it is not possible to search for the reflection element information including the utterance content from each of the reflection element information, the reflexive determination section 320 converts the topic search command signal input from the sentence structure analysis section 430 into a parallax return determination section 330. Output to
[0069]
The parrot return determination unit 330 compares the current first morpheme information extracted by the morpheme extraction unit 410 with the past answer sentence stored in the parrot return element database 802, and determines that the current first morpheme information is a past answer. If it is included in the sentence, it is a standard acquisition means for acquiring the contents of the agreement.
[0070]
Here, in this embodiment, the parrot return means that the content of the utterance of the user is repeated as it is (or a content close thereto). In this embodiment, the parrot return element is composed of first morpheme information or the like constituting an answer sentence output immediately before from the conversation control device 1, and as shown in FIG. 14, for example, "horse is beautiful"(Horse;*; beautiful) and "I like Sato"(Sato;*; I like it).
[0071]
In addition, the parrot return element database 802 is an agreement storage unit that stores agreement details for agreeing to input information input by a user in advance. The contents of the agreement include, for example, the input information previously input by the user (if the previous input information is “Is the director of the A movie name Mr. S?” The director of A movie name is Mr. S)), or "Exactly", "True".
[0072]
Specifically, the parrot return determining unit 330 to which the topic search command signal is input from the reflexive determination unit 320 configures the first morphological information and the parrot return element included in the input topic search command signal for each parrot return element. By comparing each morpheme, it is determined whether or not the first morpheme information is included in the parrot return element (see FIG. 14).
[0073]
When the parrot return determination unit 330 determines that the first morpheme information is included in each parrot return element, it obtains the consensus content and outputs an answer sentence including the obtained consensus content to the management unit 310 (parrot return). Process). That is, assuming that the parrot return element (such as the previous answer sentence) is S and the first morpheme information is W, the parrot return determination unit 330 indicates above when the relationship of W⊂S and W ≠ φ is established. Perform parrot return processing.
[0074]
For example, the conversation control device 1 gives the answer sentence "The director of the A movie name is Mr. S" (the director of the A movie name; Mr. S; *) (the order is subject, object, action, and so on). ) Is output, and then, when the user utters “A is the director of the A movie name Mr. S” (the director of the A movie name; Mr. S; *) in response to the output sentence, the parrot returns The determination unit 330 determines that the first morpheme information of the user (Director of the A movie name; Mr. S; *) matches each morpheme of the answer sentence (Director of the A movie name; Mr. S; *). The user concludes that the answer is being parroted, obtains the stored contents of the agreement, such as "exactly," and outputs the obtained agreement.
[0075]
The parrot return determining unit 330 compares the current first morphological information extracted by the morphological extracting unit 410 with the past first morphological information stored in the parrot return element database 802, and outputs the current first morphological information. Is included in the past first morpheme information, it is also a fixed form acquisition means for acquiring the content of repulsion.
[0076]
More specifically, when the user utters the utterance content “Horse is beautiful” and the conversation control device 1 outputs the content of “Horse may have a sense of dynamism” as an answer sentence, it is used later. When the person repeats the utterance content “horse is beautiful”, the parrot return determining unit 330 determines that each morpheme (first morpheme information) constituting the current utterance content “horse is beautiful” {horse; Each morpheme (first morpheme information) that constitutes the previous utterance content “Horse is beautiful” {horse; *; beautiful} matches, so the user can reply from the conversation control device 1 “Horse is dynamic. "I have a good feeling."
[0077]
In this case, since the user has not heard the answer sentence from the conversation control device 1, the parrot return determination unit 330 acquires the stored rebound content (for example, do not repeat the same content) and obtains the obtained repulsion. The contents can be output.
[0078]
On the other hand, when the parrot return determining unit 330 determines that the first morpheme information is the same as the content of the previous answer sentence or that the first morphological information is not the same as the previous first morphological information, the reflexive determining unit 320 The input topic search command signal is output to the discourse range determination unit 340.
[0079]
Note that the parrot return determination unit 330 has described the process in the case where the user has performed the parrot return on the "answer sentence of the conversation control device 1", but the following process can also be performed. For example, when the output unit 600 outputs an answer sentence “Horse is beautiful”, the user responds to this answer with “Why is the horse beautiful?”, “Why is it beautiful?”, Or “Why is it?” This is the process of the parrot return determination unit 330 performed when the user utters.
[0080]
In this case, the parrot return determination unit 330 outputs the answer sentence S “the horse is beautiful” and the utterance content W from the user (“Why is the horse beautiful? (Question sentence)” or “Why is it beautiful? (Question sentence). )), (W−c) ⊂S, S ≠ φ, c ≠ φ (where c represents the utterance type of W, and this utterance type is determined by the utterance type determination unit 440 described below. The utterance type includes, for example, a question sentence, as will be described later.) Therefore, “conditional” parrot return processing (when the user Process when a parrot with a mark is performed).
[0081]
As the “conditional” parrot return processing, for example, when the conversation control device 1 outputs an answer sentence “Horse is beautiful”, the user utters the utterance content “Why is the horse beautiful?” At this time, in order to solve the user's questions and the like, the parrot return determining unit 330 obtains an answer sentence such as “The horse is not beautiful” from the parrot return element database 802, and the obtained answer sentence is stored in the management unit 310. Perform output processing.
[0082]
The discourse range determination unit 340 compares the first morpheme information extracted by the phrase analysis unit 420 with each of the prestored discourse ranges, and matches the morpheme included in the first morpheme information from each discourse range. This is a discourse range search means for searching the discourse range.
[0083]
Specifically, the discourse range determination unit 340 to which the topic search command signal is input from the parrot return determination unit 330 determines the user's discourse range based on the input discourse search command signal. That is, the discourse range determination unit 340 searches the conversation database 500 for a range relevant to the content spoken by the user (discussion range) based on the input search command signal.
[0084]
For example, the discourse range determination unit 340 determines that the first morphological information included in the input topic search command signal is (interesting movie; *; there is) @ is there any interesting movie? In the case of｝, the first morpheme information is collated with the discourse range group, and when the discourse range group includes a morpheme (for example, “movie”) constituting the first morpheme information, The "movie" included in the information is determined as the discourse range. In this case, since the discourse range “movie” is included in the first morpheme information, the discourse range determination unit 340 includes the input first morpheme information in the topic search command signal and outputs it to the topic search unit 360.
[0085]
On the other hand, if the discourse range group is not included in the first morpheme information, the discourse range determination unit 340 includes the input first morpheme information in the topic search command signal and outputs it to the abbreviation sentence complementing unit 350.
[0086]
Thereby, the topic search unit 360, which will be described later, checks each “topic title” belonging to the “discussion range” determined by the discourse range determination unit 340 against the first morpheme information specified by the sentence structure analysis unit 430. Therefore, it is not necessary to collate “all” “topic titles” (second morpheme information) with the first morpheme information, and the answer sentence search unit 370 described later performs until the final answer sentence is searched. Time can be shortened.
[0087]
As described above, the discourse range determination unit 340 compares the first morpheme information with the discourse range group, and if the discourse range group includes the morpheme of the first morpheme information, determines the morpheme as the discourse range. However, the present invention is not limited to this, and the morpheme of the parrot return element searched immediately before by the parrot return determination unit 330 or the morpheme constituting the utterance content uttered by the user may be determined as the discourse range. . Using the discourse range determined by the discourse range determination unit 340, the abbreviated sentence complementing unit 350 described below can add the discourse range to the first morpheme information in which the morpheme is omitted.
[0088]
The abbreviated sentence complementing unit 350 searches for an attribute that does not include a morpheme from among the attributes (subject, object, action, etc.) constituting the first morpheme information based on the first morpheme information extracted by the phrase analysis unit 420. Attribute search means. The abbreviation sentence complementing unit 350 is also a morpheme adding unit that adds a morpheme constituting the discourse range searched by the discourse range determination unit 340 to the attribute based on the searched attribute.
[0089]
Specifically, the abbreviation sentence complementing unit 350 to which the topic search command signal has been input from the discourse range determination unit 340, based on the first morpheme information included in the input discourse search command signal, generates an utterance composed of the first morpheme information. It is determined whether the content is an abbreviated sentence. If the utterance content composed of the first morpheme information is an abbreviated sentence, the morpheme of the discourse range to which the first morpheme information belongs is added to the first morpheme information.
[0090]
For example, the abbreviation sentence complementing unit 350 determines that the morpheme constituting the first morpheme information included in the input topic search command signal is (supervisor; *; *) (supervisor?) Is unknown, meaning an abbreviated sentence.), The discourse range determined by the discourse range determination unit 340 (A movie name; the A movie name indicates the title of the movie) If the first morpheme information belongs to a certain morpheme information, the determined discourse range (A movie name) is added to the first morpheme information to the morphemes constituting the first morpheme information (director of “A movie name”; *; *) Yes.
[0091]
That is, assuming that the first morpheme information is W and the determined discourse range is D, the abbreviation sentence complementing unit 350 adds the discourse range D to W and includes the added first morpheme information in the topic search command signal. Output to the topic search unit 360.
[0092]
With this, even when the first morpheme information is an abbreviation and is not clear as Japanese, the abbreviation sentence complementing unit 350, for example, if the first morpheme information belongs to a certain discourse range, The discourse range D (A movie name) is added to the first morpheme information W (Director; *; *), and the first morpheme information is W '(Director of A movie name; *; *) ｛Director of the A movie name is ? Because it can be treated as｝, even if the user's utterance content is an abbreviation, the abbreviation can be complemented based on the previously determined discourse range, and the abbreviation should be clarified Can be.
[0093]
For this reason, the abbreviation sentence complementing unit 350 generates the first morpheme information so that even if the utterance content constituting the first morpheme information is an abbreviation, the utterance content constituting the first morpheme information becomes proper Japanese. The topic search unit 360 obtains an optimal “topic title” (second morpheme information) related to the first morpheme information based on the complemented first morpheme information. The answer sentence search unit 370 can output an answer sentence more suitable for the uttered content of the user based on the “topic title” acquired by the topic search unit 360.
[0094]
The topic search unit 360 collates the first morpheme information extracted by the phrase analysis unit 420 or the first morpheme information complemented by the abbreviation sentence complementing unit 350 with each second morpheme information, Search means for searching for second morpheme information including morphemes constituting the first morpheme information from the inside.
[0095]
Specifically, the topic search unit 360 to which the topic search command signal has been input from the discourse range determination unit 340 or the abbreviation sentence complementing unit 350 has the discourse range based on the first morpheme information included in the input topic search command signal. From each "topic title" (second morpheme information) belonging to the discourse range determined by the determination unit 340, a "topic title" including the morpheme of the first morpheme information is searched, and this search result is used as a search result signal. Output to answer sentence search unit 370 and utterance type determination unit 440.
[0096]
For example, if the “case configuration” that constitutes the first morphological information is (Sato; *; like) {I like Sato}, the topic search unit 360 determines the “case” as shown in FIG. The morphemes (Sato; *; I like) belonging to “composition” are compared with the topic titles 1-1 to 1-4 belonging to the discourse range (Sato), and from among the topic titles 1-1 to 1-4, A topic title 1-1 (Sato; *; I like) that matches (or approximates) each morpheme (Sato; *; I like) belonging to "case composition" is searched, and the search result is used as an answer sentence as a search result signal. The data is output to the search unit 370 and the utterance type determination unit 440.
[0097]
The utterance type determination unit 440 to which the search result signal is input from the topic search unit 360 is an answer search command for causing the user to search for a specific answer sentence based on the input search result signal. The signal (the answer search command signal also includes the determined “uttered sentence type”) is output to the answer sentence search unit 370.
[0098]
The answer sentence search unit 370 is an answer search unit that searches for an answer sentence associated with the second morpheme information based on the second morpheme information (topic title) searched by the topic search unit 360. Also, the answer sentence search unit 370 collates the utterance type of the identified user with each answer type associated with the second morpheme information based on the second morpheme information searched by the topic search unit 360, It is also a second search means for acquiring an answer type that matches the utterance type of the user from among the answer types.
[0099]
Specifically, the answer sentence search unit 370 to which the search result signal from the topic search unit 360 and the answer search command signal from the utterance type determination unit 440 have been input, outputs the topic title (search result) corresponding to the input search result signal. Based on the second morpheme information) and the "utterance sentence type" (utterance type) corresponding to the answer search instruction signal, from among the answer sentence group associated with the "topic title", An answer sentence having an answer type (this answer type means "answer sentence type" shown in FIG. 11) matching the "sentence type" (DA, IA, CA, etc.) is searched.
[0100]
For example, if the topic title corresponding to the search result is the topic title 1-1 (Sato; *; I like it) shown in FIG. 12, the answer sentence search unit 370 is associated with the topic title 1-1. From the answer sentence 1-1 (DA, IA, CA, etc.), an answer consisting of an answer type (DA) that matches the "utterance sentence type" (for example, DA; utterance type) determined by the utterance type determination unit 440 The sentence 1-1 (DA; (I also like Sato)) is searched, and the searched answer sentence is output to the management unit 310 as an answer sentence signal.
[0101]
The management unit 310 to which the answer sentence signal is input from the answer sentence search unit 370 outputs the input answer sentence signal to the output unit 600. In addition, the management unit 310 to which the reflection element information is input from the reflective judgment unit 320 or the content of the parrot return processing from the parrot return determination unit 330 is input, and the answer sentence corresponding to the input reflection element information, the content of the input parrot return processing Is output to the output unit 600.
[0102]
The output unit 600 is an output unit that outputs the answer sentence obtained by the answer sentence search unit 370. In the present embodiment, for example, a speaker, a display, or the like is used. Specifically, the output unit 600 to which the answer sentence is input from the management unit 310 outputs the input answer sentence {for example, I also like Sato}.
[0103]
(Conversation control method using conversation control device)
The conversation control method by the conversation control device 1 having the above configuration can be implemented by the following procedure. FIG. 15 is a flowchart illustrating a procedure of the conversation control method according to the present embodiment.
[0104]
First, the input unit 100 performs a step of acquiring the utterance content from the user (S101). Specifically, the input unit 100 acquires a voice that constitutes the utterance content of the user, and outputs the acquired voice to the voice recognition unit 200 as a voice signal. Further, the input unit 100 specifies a character string included in the input information (other than voice) based on the input information (other than voice) input from the user, and uses the specified character string as a character string signal as a conversation controller. Output to 300.
[0105]
Next, the voice recognition unit 200 performs a step of specifying a character string included in the utterance content based on the utterance content acquired by the input unit 100 (S102). Specifically, the voice recognition unit 200 to which the voice signal is input from the input unit 100 analyzes the input voice signal, and stores a character string included in the analyzed voice signal in the voice recognition dictionary storage unit 700. The specified character string is output to the conversation control unit 300 as a character string signal.
[0106]
Then, the morpheme extraction unit 410 performs a step of extracting each morpheme constituting the minimum unit of the character string as the first morpheme information based on the character string specified by the speech recognition unit 200 (S103).
[0107]
Specifically, the morpheme extraction unit 410 to which the character string signal has been input from the management unit 310 includes a character string corresponding to the input character string signal and a noun, adjective, verb, or the like stored in the morphological database 450 in advance. The morphemes are collated with each other, each morpheme (m1, m2,...) That matches the morpheme group is extracted from the character string, and each extracted morpheme is output to the phrase analyzer 420 as an extraction signal.
[0108]
Then, the phrase analysis unit 420 performs a step of compiling each morpheme into a phrase format based on each morpheme extracted by the morpheme extraction unit 410 (S104). Specifically, the phrase analysis unit 420 to which the extraction signal is input from the morpheme extraction unit 410 uses the respective morphemes corresponding to the input extraction signal to combine them into a phrase format.
[0109]
That is, as shown in FIG. 4, the phrase analysis unit 420 extracts a dependency element (for example, ga wo wo ...) of each morpheme based on each morpheme corresponding to the input extracted signal and extracts the morpheme. Each morpheme is grouped into each clause based on the determined dependency element. The phrase analysis unit 420 that compiles each morpheme into each phrase, the sentence structure analysis unit 430 and the utterance type determination unit 440, using the morphological information including each morpheme that summarizes each morpheme and each morpheme that composes each phras as a morphological signal. Output to
[0110]
After that, the sentence structure analysis unit 430 performs a step of classifying each morpheme of the first morpheme information segmented by the phrase analysis unit 420 into attributes such as subject matter and object case (S105). Specifically, the sentence structure analysis unit 430 to which the sentence pattern signal is input from the sentence pattern analysis unit 420, based on each morpheme corresponding to the input sentence pattern signal and the sentence formed by each morpheme, Determine the "case composition".
[0111]
That is, as shown in FIG. 5, for example, when the dependency element of each morpheme is "" or "ha", the sentence structure analysis unit 430 determines that the morpheme before the dependency element is the subject (subject) Or nominative). In addition, for example, when the dependency element of each morpheme is "" or "", the sentence structure analysis unit 430 determines that the morpheme before the dependency element is an object (object).
[0112]
Further, for example, when the dependency element of each morpheme is “do”, the sentence structure analysis unit 430 determines that the morpheme preceding the dependency element is an action (predicate; this predicate is composed of a verb, an adjective, etc.). Is determined.
[0113]
The sentence structure analysis unit 430 that has determined the “case configuration” of each morpheme constituting each phrase, based on the first morpheme information associated with the determined “case configuration”, determines the range of a topic described below. A topic search command signal for specifying is output to topic search section 360.
[0114]
Next, the utterance type determination unit 440 performs a step of specifying an utterance type indicating the type of utterance content based on the phrase specified by the phrase analysis unit 420 (S106). Specifically, the “speech sentence type” (speech type) is determined based on each morpheme corresponding to the sentence pattern signal input from the phrase analysis unit 420 and the clause composed of each morpheme.
[0115]
That is, based on each phrase corresponding to the input sentence pattern signal, the utterance type determination unit 440 collates each phrase with each dictionary stored in the utterance type database 460, and from among each phrase, Extract sentence elements related to the dictionary. The utterance type determination unit 440 that extracts a sentence element related to each dictionary from each phrase determines a “speech sentence type” based on the extracted sentence element.
[0116]
The utterance type determination unit 440 outputs an answer search command signal for causing the corresponding user to search for a specific answer sentence to the answer sentence search unit 370 based on an instruction from the topic search unit 360 described later.
[0117]
Next, the reflexive determination unit 320 performs a step of comparing the first morpheme information extracted by the morpheme extraction unit 410 with each of the fixed contents, and searching for the fixed contents including the first morpheme information from among the fixed contents. (S107: reflective processing).
[0118]
Specifically, the reflexive determination unit 320 to which the topic search command signal has been input from the sentence structure analysis unit 430 receives the first morphological information included in the input topic search command signal and each of the information stored in the reflection element database 801. The reflection element information is compared with the reflection element information (standard content), the reflection element information including the first morpheme information is searched from each reflection element information, and the searched reflection element information is output to the management unit 310.
[0119]
When it is not possible to search for the reflection element information including the first morpheme information from among the respective pieces of reflection element information, the reflexive determination unit 320 determines the topic search command signal input from the sentence structure analysis unit 430 in a parable manner. Output to the unit 330.
[0120]
Next, the parrot return determining unit 330 performs a step of comparing the first morphological information extracted by the morphological extraction unit 410 with each parrot return element, and searching for a parrot return element including the first morphological information from each parrot return element ( S108: parrot return processing).
[0121]
When the parrot return determining unit 330 determines that the first morpheme information is included in each parrot return element, the parrot return element acquires the parrot return element including the first morphological information, and manages an answer sentence including the obtained parrot return element. Output to the unit 310 (parrot return processing). That is, if the parrot return element (response sentence output last time, the utterance content of the previous user's utterance, etc.) is S and the first morpheme information is W, the parrot return determination unit 330 determines that the relationship of W⊂S and W ≠ φ If it is established, the parrot return processing described above is performed.
[0122]
On the other hand, if the parrot return determining unit 330 determines that the first morphological information is not included in each parrot return element, the parrot return determining unit 330 sends the topic search command signal input from the reflective determining unit 320 to the discourse range determining unit 340. Output.
[0123]
Then, the discourse range determination unit 340 compares the first morpheme extracted by the phrase analysis unit 420 with each discourse range, and searches (determines) a discourse range including the first morpheme information from each discourse range. The step is performed (S109).
[0124]
Specifically, the discourse range determining unit 340 to which the topic search command signal has been input from the parrot return determination unit 330 determines, based on the input search command signal, the contents of the conversation database 500 that the user is speaking. Search for a relevant range (discussion range).
[0125]
For example, the discourse range determination unit 340 determines that the first morphological information included in the input topic search command signal is (interesting movie; *; there is) @ is there any interesting movie? In the case of｝, the first morpheme information is collated with the discourse range group, and when the discourse range group includes a morpheme (for example, “movie”) constituting the first morpheme information, The "movie" included in the information is determined as the discourse range. In this case, since the discourse range “movie” is included in the first morpheme information, the discourse range determination unit 340 includes the input first morpheme information in the topic search command signal and outputs it to the topic search unit 360.
[0126]
On the other hand, if the discourse range group is not included in the first morpheme information, the discourse range determination unit 340 includes the input first morpheme information in the topic search command signal and outputs it to the abbreviation sentence complementing unit 350.
[0127]
Next, the abbreviation sentence complementing section 350 selects, from among the attributes (subject, object, action, etc.) constituting the first morphological information based on the first morphological information extracted by the phrase analyzing section 420, an attribute that does not include a morpheme. Perform the step of searching for Thereafter, the abbreviation sentence complementing unit 350 performs a step of adding a morpheme constituting the discourse range searched by the discourse range determination unit 340 to the attribute based on the attribute that does not include the searched morpheme (S110; abbreviation sentence). Complement).
[0128]
Specifically, the abbreviation sentence complementing unit 350 to which the topic search command signal has been input from the discourse range determination unit 340, based on the first morpheme information included in the input discourse search command signal, generates an utterance composed of the first morpheme information. It is determined whether the content is an abbreviated sentence. If the utterance content composed of the first morpheme information is an abbreviated sentence, the morpheme of the discourse range to which the first morpheme information belongs is added to the first morpheme information.
[0129]
For example, the abbreviation sentence complementing unit 350 determines that the morpheme constituting the first morpheme information included in the input topic search command signal is (supervisor; *; *) (supervisor?) Is unknown, so it means an abbreviation.) In the case of, the discourse range previously determined by the discourse range determination unit 340 (A movie name; the A movie name is the movie title) If the first morpheme information belongs to the first morpheme information, the morpheme of the determined discourse range (A movie name) is added to the first morpheme information (“A movie name”). *; *).
[0130]
That is, assuming that the first morpheme information is W and the determined discourse range is D, the abbreviation sentence complementing unit 350 adds the discourse range D to the first morpheme information W, and uses the post-addition first morpheme information as a topic search command. It is included in the signal and output to the topic search unit 360.
[0131]
Next, the topic search unit 360 collates the first morpheme information extracted by the phrase analysis unit 420 or the first morpheme information complemented by the abbreviation sentence complementing unit 350 with each second morpheme information, A step of searching the information for second morpheme information including morphemes constituting the first morpheme information is performed (S111).
[0132]
Specifically, the topic search unit 360 to which the topic search command signal has been input from the discourse range determination unit 340 or the abbreviation sentence complementing unit 350 has the discourse range based on the first morpheme information included in the input topic search command signal. From each "topic title" (second morpheme information) belonging to the discourse range determined by the determination unit 340, a "topic title" including the morpheme of the first morpheme information is searched, and this search result is used as a search result signal. Output to answer sentence search unit 370 and utterance type determination unit 440.
[0133]
For example, if the “case configuration” that constitutes the first morphological information is (Sato; *; like) {I like Sato}, the topic search unit 360 determines the “case” as shown in FIG. The morphemes (Sato; *; I like) belonging to “composition” are compared with the topic titles 1-1 to 1-4 belonging to the discourse range (Sato), and from among the topic titles 1-1 to 1-4, A topic title 1-1 (Sato; *; I like) that matches (or approximates) each morpheme (Sato; *; I like) belonging to "case composition" is searched, and the search result is used as an answer sentence as a search result signal. The data is output to the search unit 370 and the utterance type determination unit 440.
[0134]
The utterance type determination unit 440 to which the search result signal is input from the topic search unit 360 is an answer search command for causing the user to search for a specific answer sentence based on the input search result signal. The signal (the answer search command signal also includes the determined “uttered sentence type”) is output to the answer sentence search unit 370.
[0135]
Then, the answer sentence search unit 370 collates the utterance type of the specified user with each answer type associated with the second morpheme information based on the second morpheme information searched by the topic search unit 360, From each answer type, a step of searching for an answer type that matches the utterance type of the user and searching for an answer sentence associated with the answer type based on the searched answer type is performed (S112).
[0136]
Specifically, the answer sentence search unit 370 to which the search result signal from the topic search unit 360 and the answer search command signal from the utterance type determination unit 440 have been input, outputs the topic title (search result) corresponding to the input search result signal. Based on the second morpheme information) and the "utterance sentence type" (utterance type) corresponding to the answer search instruction signal, from among the answer sentence group associated with the "topic title", An answer sentence having an answer type (this answer type means "answer sentence type" shown in FIG. 11) matching the "sentence type" (DA, IA, CA, etc.) is searched.
[0137]
For example, if the topic title corresponding to the search result is the topic title 1-1 (Sato; *; I like it) shown in FIG. 12, the answer sentence search unit 370 is associated with the topic title 1-1. From the answer sentence 1-1 (DA, IA, CA, etc.), an answer consisting of an answer type (DA) that matches the "utterance sentence type" (for example, DA; utterance type) determined by the utterance type determination unit 440 The sentence 1-1 (DA; (I also like Sato)) is searched, and the searched answer sentence is output to the management unit 310 as an answer sentence signal.
[0138]
Next, the management unit 310 to which the answer sentence signal has been input from the answer sentence search unit 370 outputs the input answer sentence signal to the output unit 600. In addition, the management unit 310 to which the reflection element information is input from the reflective judgment unit 320 or the content of the parrot return processing from the parrot return determination unit 330 is input, and the answer sentence corresponding to the input reflection element information, the content of the input parrot return processing Is output to the output unit 600 (S113). The output unit 600 to which the answer sentence is input from the management unit 310 outputs the input answer sentence {for example, I also like Sato}.
[0139]
(Operation and effect of the conversation control device and the conversation control method)
According to the invention of the present application having the above-described configuration, the reflexive determination unit 320 collates the first morpheme information constituting the utterance content uttered by the user with each of the pre-stored standard contents, and from among, it is possible to find the standard content including a first morpheme information, reflexive determination unit 320, for example when the first morpheme information is a fixed content such as "Hello" includes the standard content it is possible to answer the same boilerplate content "Hello" or the like.
[0140]
When the utterance content of the user is a fixed content, the reflexive determination unit 320 answers the fixed content (such as greeting), so that the user first communicates with the conversation control device 1. You can enjoy the feeling of communicating.
[0141]
In addition, the parrot return determining unit 330 compares the current first morpheme information with the past answer sentence, and if the current first morpheme information is not included in the past answer sentence, an agreement stored in advance. Since the content can be acquired, if the input information currently input by the user matches the past answer sentence, the parrot return determining unit 330 returns the user to the past answer sentence That the user has listened to the answer sentence again).
[0142]
In this case, the parrot return determination unit 330 obtains the stored agreement contents because the user has performed a parrot with respect to the past answer sentence, and obtains the obtained agreement contents (for example, “yes”). Can be output. Thus, if the user does not understand the meaning of the answer sentence output from the conversation control device 1, the user can listen to the answer again and listen to the answer again, so that the user is just talking with another user. You can enjoy such feeling.
[0143]
In addition, the parrot return determining unit 330 compares the current first morpheme information with the past first morpheme information, and obtains the repulsion content when the current first morpheme information is included in the past first morpheme information. If the input information input last time is included in the input information input this time, the user repeatedly inputs the same content as the previous input information. It can be determined that the user has not responded properly to the answer sent from the conversation control device.
[0144]
In this case, the parrot return determination unit 330 obtains the stored repulsion content to repel the user because the user has not properly responded to the previous answer sentence, and obtains the obtained repulsion content. Is output. Thereby, unless the user inputs appropriate input information for the answer sentence from the conversation control device 1, the content of the repulsion is output from the conversation control device 1, so that the user can talk with another user. You can enjoy the feeling of being.
[0145]
Also, the topic search unit 360 may search each second morpheme information belonging to the “discussion range” with the first morpheme information to search for second morpheme information that is similar to the first morpheme information. It is not necessary to collate the second morpheme information with the first morpheme information, so that the time required to search for second morpheme information similar to the first morpheme information can be reduced.
[0146]
As a result, the topic search unit 360 can quickly search (pinpoint search) for the second morpheme information that is similar to the first morpheme information, so that the answer sentence search unit 360 The answer sentence associated with the second morpheme information can be acquired in a short time based on the retrieved second morpheme information, and the conversation control device 1 quickly answers the utterance content from the user. be able to.
[0147]
Also, the topic search unit 360 searches the second morpheme information for second morpheme information including morphemes constituting the first morpheme information (elements constituting the utterance content of the user), and an answer sentence search unit Since the 370 can search for an answer sentence associated with the second morpheme information based on the second morpheme information searched by the topic search unit 360, the answer sentence search unit 370 searches for the utterance content of the user. Based on the constituent morphemes (first morpheme information), a semantic space (subject, target, etc.) constructed by each morpheme can be considered, and an answer sentence created in advance based on the semantic space can be searched. Thus, it is possible to search for an answer sentence that is more suitable for the utterance content than simply using the entire utterance content as a keyword and searching for an answer sentence associated with that keyword.
[0148]
Further, since the topic search unit 360 searches for the second morpheme information including the first morpheme information, there is no need to search for the second morpheme information that completely matches the utterance content of the user, and the conversation control device 1 is developed. A developer who does not need to previously store an enormous amount of answer sentences corresponding to the utterance contents likely to be uttered by the user, and can reduce the capacity of the storage unit.
[0149]
Further, the answer sentence search unit 370 selects an answer type that matches the utterance type of the user from among the answer types (statements, affirmations, places, repulsions, etc.) associated with each second morpheme information belonging to the “discussion range”. Can be searched for, and the answer sentence associated with the answer type can be searched based on the searched answer type. Therefore, the conversation control device 1 Based on a simple statement, an impression of the user, or a description of the locational element of the user, an answer sentence that matches the utterance type of the user is selected from multiple answer sentences. The search can be performed, and a more optimal answer can be given to the corresponding user.
[0150]
Furthermore, the answer sentence search unit 370 selects, from among the answer types associated with each second morpheme information belonging only to the “discussion range” retrieved by the discourse range determination unit 340, an answer that matches the utterance type of the user. Since it is only necessary to search for the type (pinpoint search is possible), there is no need to search for the answer type and the user's utterance type associated with the "all" second morpheme information one by one, and the user's utterance type Can be obtained in a short time.
[0151]
Finally, the abbreviation sentence complementing unit 350 determines that the first morpheme information belonging to the discourse range where the first morpheme information is present even if the first morpheme information constituting the utterance content of the user is an abbreviation and is not clear as Japanese. If so, the discourse range can be added to the first morpheme information to complement the first morpheme information composed of abbreviations.
[0152]
With this, the abbreviation sentence complementing unit 350 generates the first morpheme information so that the utterance content constituting the first morpheme information becomes proper Japanese even if the utterance content constituting the first morpheme information is an abbreviation sentence. Can be supplemented with a specific morpheme (such as a morpheme constituting the discourse range), the topic search unit 360 determines the first morpheme based on the complemented first morpheme information supplemented by the abbreviation sentence complementing unit 350. The optimal second morpheme information related to the information can be acquired, and the answer sentence search unit 370 can select an answer sentence more suitable for the uttered content of the user based on the second morpheme information acquired by the topic search unit 360. Can be output.
[0153]
As a result, even if the input information from the user is an abbreviated sentence, the conversation control device 1 uses the past search result without using the functions such as the neural network and the AI intelligence to determine what the abbreviated sentence means. It is possible to infer whether the conversation control device 1 is to be used, and the developer of the conversation control device 1 does not need to install a neural network and AI intelligence, so that the system of the conversation control device 1 can be constructed more easily.
[0154]
[Example of change]
It should be noted that the present invention is not limited to the above embodiment, and the following changes can be made.
[0155]
(First modification example)
In the present modification, the conversation database 500 may be an element storage unit that stores a plurality of element information indicating the entire set group including a plurality of morpheme sets in association with the set group. Further, the morpheme extraction unit 410 compares the morpheme extracted from the character string with each set group, selects a set group including the extracted morpheme from each set group, and selects an element associated with the selected set group. Information may be extracted as first morpheme information.
[0156]
As shown in FIG. 16, some morphemes included in a character string uttered by the user are similar. For example, as shown in FIG. 16, when the element information indicating the entire set group is “gift”, the “gift” is similar to a present, a gift, a year-end gift, a gift, a celebration, etc. (set group). Therefore, if there is a morpheme similar to “gift” (such as the above-described present), the morpheme extraction unit 410 can treat the similar morpheme as “gift”.
[0157]
That is, for example, when the morpheme extracted from the character string is “present”, the morpheme extracting unit 410 determines that the element information representing “present” is “gift” as shown in FIG. "Present" can be replaced with "gift".
[0158]
This allows the morpheme extraction unit 410 to sort morphemes similar to each other, so that a developer who develops a conversation control device can use the morpheme extraction unit 410 corresponding to the semantic space grasped from each of the mutually similar first morpheme information. It is not necessary to create answer sentences related to the two morpheme information and the second morpheme information one by one, and as a result, the amount of data to be stored in the storage unit can be reduced.
[0159]
(Second modification example)
As shown in FIG. 17, in this modification, the topic search unit 360 may include a ratio calculation unit 361 and a selection unit 362.
[0160]
The ratio calculator 361 compares the first morpheme information extracted by the morpheme extractor 410 with each second information, and for each second morpheme information, a ratio of the first morpheme information to the second morpheme information. Is a calculation means for calculating.
[0161]
Specifically, as shown in FIG. 17, the ratio calculation unit 361 to which the topic search command signal is input from the sentence structure analysis unit 430, generates a second morpheme information based on the first morpheme information included in the input topic search command signal. The one morpheme information is compared with each topic title (second morpheme information) belonging to the discourse range stored in the conversation database 500, and for each topic title, the first morpheme information occupies each topic title. Calculate the percentage.
[0162]
For example, as shown in FIG. 18, when the first morpheme information constituting the utterance sentence uttered by the user is (Sato; *; I like) {I like Sato}, the ratio calculation unit 361 Each morpheme (Sato; *; I like) belonging to "case composition" is compared with each morpheme (Sato; *; I like) included in the topic title, and each morpheme belonging to "case composition" is added to the topic title. (Sato; *; like) is calculated as 100%. The ratio calculation unit 361 performs these calculations for each topic title, and outputs the calculated ratios to the selection unit 362 as ratio signals.
[0163]
The selection unit 362 is a selection unit that selects one piece of second morpheme information from each piece of second morpheme information according to the magnitude of each ratio calculated for each piece of second morpheme information by the ratio calculation unit 361. is there.
[0164]
Specifically, the selection unit 362 to which the ratio signal is input from the ratio calculation unit 361 selects from among the ratios (element of “case composition” / element of “topic title” × 100) included in the input ratio signal. For example, a topic title having a high ratio is selected (see FIG. 18). The selecting unit 362 that has selected the topic title having a high ratio outputs the selected topic title to the answer sentence searching unit 370 and the utterance type determining unit 440 as a search result signal. The answer sentence search unit 370 searches for an answer sentence associated with the topic title based on the topic title selected by the selection unit 362.
[0165]
Accordingly, the selecting unit 362 calculates, for each piece of the second morpheme information, the ratio of the first morpheme information to the second morpheme information, and calculates the magnitude of each ratio calculated for each piece of the second morpheme information. The second morpheme information can be selected from among the pieces of second morpheme information in accordance with the above. Therefore, the selecting unit 362 determines that the first morpheme information (which constitutes the utterance content of the user) If the second morpheme information having a large proportion in the second morpheme information can be obtained from a plurality of second morpheme information groups, the second morpheme information following the semantic space grasped from the first morpheme information can be more efficiently obtained. As a result, the answer sentence search unit 370 can give an optimal answer to the utterance content of the user.
[0166]
Further, since the selecting unit 362 can select a topic title having a high ratio calculated by the ratio calculating unit 361 from a plurality of topic titles, the selecting unit 362 belongs to the “case configuration” included in the utterance sentence of the user. Even if each morpheme and each topic title stored in the conversation database 500 do not completely match, it is possible to acquire a topic title closely related to each morpheme belonging to “case composition”.
[0167]
As a result, the selecting unit 362 can acquire a topic title closely related to the “case configuration” that forms the first morpheme information, so that the developer who develops the conversation control device 1 configures “ It is not necessary to store topic titles that completely match the "case configuration" in the conversation database 500 one by one, so that the capacity of the conversation database 500 can be reduced.
[0168]
Further, the ratio calculation unit 361 calculates the ratio of the first morpheme information to the second morpheme information for each second morpheme information belonging only to the “discussion range” retrieved by the discourse range determination unit 340. Therefore, there is no need to calculate the ratio of the first morpheme information to all the second morpheme information, and the second morpheme information following the semantic space composed of the first morpheme information is obtained in a shorter time. As a result, based on the acquired second morphological information, it is possible to quickly output an optimal answer sentence to the utterance content from the user.
[0169]
Note that the ratio calculation unit 361 collates each morpheme of the first morpheme information belonging to each classified attribute with each morpheme of the second morpheme information belonging to each attribute stored in advance for each attribute. The first morpheme information may be a first retrieval unit that retrieves second morpheme information that includes each morpheme of the first morpheme information in at least one attribute from the second morpheme information.
[0170]
Specifically, the ratio calculation unit 361 to which the topic search command signal has been input, for each “case” (subject; object; action) of the “case configuration” included in the input topic search command signal, generates the “case”. Is compared with each morpheme belonging to the topic title “case” having the same “case”, and it is determined whether or not the morphemes constituting the “case” are the same.
[0171]
For example, as shown in FIG. 19, when the morpheme of the “case” of the “case configuration” is (dog; person; bites) {the dog bites the person}, the ratio calculation unit 361 Match "dog", "person", "bite" with the morpheme "dog", "person", "bite" of the topic title consisting of "case" which is the same as "case" that composes those morphemes Then, among the morphemes “dog”, “person”, and “bite” constituting the topic title, the morpheme “dog” of “case composition” consisting of the same “case” as the “case” corresponding to each morpheme, Calculate (100%) the percentage that matches “person” and “bite”.
[0172]
If the element composing the topic title is (person; dog; bite) {the person bites the dog}, the ratio calculation unit 361 performs the morpheme belonging to the two cases by the same procedure as described above. Therefore, the degree of matching between the morphemes constituting the “case composition” and the “topic title” for each “case” is calculated to be 33% (see FIG. 19).
[0173]
The ratio calculation unit 361 that has calculated the ratio selects a topic title having a high ratio from among the ratios, and outputs the selected topic title to the answer sentence search unit 370 and the utterance type determination unit 440 as a search result signal.
[0174]
Accordingly, the ratio calculation unit 361 divides each morpheme of the first morpheme information belonging to each classified “case composition” (subject case, target case, etc.) and a topic title stored in advance for each “case”. Since the collation can be performed and the second morpheme information including each morpheme of the first morpheme information in at least one “case” can be searched from each topic title, the ratio calculation unit 361 differs from the normal word order. In the case of utterance contents composed of things, for example, "a person bites a dog", the subject morpheme is a "person" and the target morpheme is a "dog". The matching second morpheme information can be searched, and the answer sentence {"really?" Or "I do not understand the meaning" associated with the second morpheme information (person; dog; chew) is obtained. can do.
[0175]
That is, the ratio calculation unit 361 can identify the utterance content that is difficult to identify, for example, “people bite a dog” and “dog bite a human”. For the former, for example, "really?" For the latter, for example, "OK?"
[0176]
Further, the ratio calculation unit 361 may search the second morpheme information that includes the morpheme of the first morpheme information in at least one attribute from each of the second morpheme information belonging to the “discussion range”, and thus “all”. Of the second morpheme information, it is not necessary to obtain one second morpheme information, it is possible to obtain the second morpheme information following the semantic space composed of the first morpheme information in a shorter time, As a result, the conversation control device 1 can quickly output an optimal answer sentence to the utterance content from the user based on the acquired second morpheme information.
[0177]
Note that the selection unit 362 may select one topic title from among the topic titles according to a predetermined priority. The priority means a priority for being selected as a topic title. This priority is determined in advance by the developer at the development stage.
[0178]
(Third modification example)
As shown in FIG. 20, the present modified example is not limited to the above-described embodiment and each modified example, and transmits and receives data to and from the communication unit 800 in the conversation control devices 1a and 1b via the communication network 1000. May be provided, a conversation database 500b to 500d connected to the communication unit 900, and servers 2a to 2c (conversation control system).
[0179]
Here, the communication network 1000 means a communication network for transmitting and receiving data, and in the present embodiment, for example, the Internet is used. In this modification, the conversation databases 500b to 500d and the servers 2a to 2c are limited for convenience. However, the present invention is not limited to this, and another conversation database may be provided.
[0180]
Thereby, the conversation control unit 300 refers not only to the conversation database 500a disposed inside the conversation control device 1a but also to the other conversation control devices 1b, the other conversation databases 500b to 500d, and the servers 2a to 2c. Therefore, for example, it is not possible to search the conversation database 500a for a discourse range related to each morpheme (first morpheme information) belonging to the "case configuration" included in the topic search command signal. However, by referring to the other conversation control device 1b, the conversation databases 500b to 500d, and the servers 2a to 2c, it is possible to search the discourse range related to the "case configuration" and the like, and the A suitable answer sentence can be searched.
[0181]
(Fourth modification)
The sentence structure analysis unit 430 may store, in the conversation database 500, each “case configuration” constituting the identified first morphological information and each morpheme associated with each “case configuration”. The answer sentence search unit 370 may store, in the conversation database 500, each “case configuration” constituting the acquired answer sentence and each morpheme associated with each “case configuration”.
[0182]
The discourse range determination unit 340 may store the searched discourse range in the conversation database 500. The topic search unit 360 may store the searched second morphological information in the conversation database 500.
[0183]
The first morpheme information, the second morpheme information, each "case configuration" constituting the first morpheme information or the second morpheme information and each morpheme associated with each "case configuration", and the retrieved answer sentence The morphemes associated with each “case configuration” and each “case configuration” and the searched discourse range are associated with each other and stored in the conversation database 500 or the parrot return element database 802 as history morpheme information. be able to.
[0184]
The abbreviation sentence complementing unit 350 does not include a morpheme from each of the attributes (subject, object, action, etc .; case composition) constituting the first morphological information based on the first morphological information extracted by the phrase analyzing unit 420. The attribute is searched, and the history morpheme information stored in the conversation database 500 or the parrot return element database 802 is added to the attribute based on the searched attribute.
[0185]
Specifically, the abbreviation sentence complementing unit 350 to which the topic search command signal has been input from the discourse range determination unit 340, based on the first morpheme information included in the input discourse search command signal, generates an utterance composed of the first morpheme information. It is determined whether the content is an abbreviated sentence, and if the utterance content composed of the first morpheme information is an abbreviated sentence (for example, the subject, object, or action does not have a predetermined morpheme), the conversation database 500 or The history form information stored in the parrot return element database 802 is added to the first morpheme information.
[0186]
That is, if the subject included in the history form information is S1, the object is O1, the action A1, the discourse range is D1, and the omitted first morpheme information is W, the complemented first morpheme information W1 is S1 @ W , O1 @ W, A1 @ W, or D1 @ W.
[0187]
The topic search unit 360 collates the first morpheme information W1 complemented by the abbreviation sentence complementing unit 350 with each piece of second morpheme information, and selects the first morpheme information from each “topic title” (second morpheme information). The second morpheme information including W1 is searched, and the searched topic title is output as a search result signal to the answer sentence search unit 370 and the utterance type determination unit 440.
[0188]
Thus, even when the utterance content composed of the first morpheme information is an abbreviation and is not clear as Japanese, the abbreviation complementing unit 350 uses the history morphology information stored in the conversation database 500 to Since the morpheme of the omitted first morpheme information can be complemented, the utterance content composed of the omitted first morpheme information can be clarified.
[0189]
For this reason, if the utterance content constituting the first morpheme information is an abbreviation, the abbreviation sentence complementing unit 350 sets the first morpheme information so that the utterance content composed of the first morpheme information becomes proper Japanese. Can be complemented, the topic search unit 360 determines the optimal “topic title” (second morpheme) related to the first morpheme information based on the first morpheme information with the complemented morpheme. Information), and the answer sentence search unit 370 can output an answer sentence more suitable for the utterance content of the user based on the optimal “topic title” searched by the topic search unit 360. .
[0190]
(Fifth modification example)
The topic search unit 360 may include a deletion unit 363 and a discourse addition unit 364, as shown in FIG. The deletion unit 363 compares the second morpheme information with the discourse range retrieved by the discourse range determination unit 340 based on the retrieved second morpheme information, and selects a discourse from among the morphemes constituting the second morpheme information. This is a deletion unit that deletes a morpheme that matches the range.
[0191]
Specifically, the topic search unit 360 to which the topic search command signal is input from the abbreviation sentence complementing unit 350 receives the first morpheme information included in the input topic search command signal and the discourse determined by the discourse range determination unit 340. Each second morpheme information belonging to the range is collated, and second morpheme information that matches the first morpheme information is searched from each second morpheme information.
[0192]
Then, based on the retrieved second morpheme information, the deletion unit 363 collates the second morpheme information with the morphemes constituting the discourse range determined by the discourse range determination unit 340, and , The morpheme that matches the morpheme constituting the discourse range is deleted, and the second morpheme information from which the morpheme is deleted is output to the discourse adding unit 364 as a deletion signal.
[0193]
That is, the deletion unit 363 removes (removes) the current discourse range D2 (this D2 is made up of morphemes) determined by the discourse range determination unit 340 from each morpheme t1 constituting the second morpheme information. Assuming that the result is t2, t2 = t1-D2).
[0194]
The discourse adding unit 364 obtains another discourse range associated with the discourse range searched by the discourse range determination unit 340 based on the second morpheme information from which the morpheme has been deleted by the deletion unit 363, and obtains another obtained discourse range. This is a discourse adding means for adding morphemes constituting the discourse range to the second morpheme information.
[0195]
Specifically, assuming that the discourse range in which the current discourse range D2 is related to the answer sentence K1 is DK, another discourse having a relevance (one having a sibling relationship) to the answer sentence K1 or the current discourse range D2. Since the range D3 can be expressed as D3 = D2∪DK, the second morpheme information W2 after adding the morphemes constituting the other discourse range D3 can be expressed as W2 = t2∪D3.
[0196]
For example, each morpheme t1 constituting the second morpheme information is (A movie name; *; interesting) ｛Is the A movie name interesting? If the current discourse range D2 determined by the discourse range determination unit 340 is (A movie name), the deletion unit 363 first determines from each morpheme t1 (A movie name; *; interesting). The discourse range D2 (A movie name) is deleted, and the deleted result is set as t2 (*; *; interesting) (t2 = t1-D2).
[0197]
If the other discourse range D3 related to the current discourse range D2 (A movie name) is “B movie name”, the second morpheme information after adding the morphemes constituting the other discourse range D3 Since W2 is t2∪D3, (B movie name; *; interesting) ｛B movie name is interesting? ｝.
[0198]
Accordingly, when the user's utterance content is “A movie name is interesting?”, The discourse adding unit 364 matches each morpheme (A movie name; *; interesting) constituting the user's utterance content. The second morpheme information (A movie name; *; interesting), and the other second morpheme information (B movie name; *; interesting)？ Is the B movie name interesting? Since it can be changed to｝, the answer sentence search unit 370 acquires the answer sentence (for example, “B movie name is interesting”) associated with the second morpheme information changed by the discourse adding unit 364, The obtained answer sentence can be output.
[0199]
As a result, the answer sentence search unit 370 does not directly output an answer sentence to the utterance content of the user, but, based on the second morpheme information including the morpheme added by the discourse adding unit 364, the utterance content Can be output, and the output unit 600 can output a more personalized answer sentence based on the answer sentence searched by the answer sentence search unit 370.
[0200]
Note that the discourse adding unit 364 is not limited to the one that adds another discourse range to the second morpheme information from which the morpheme has been deleted, and the history morpheme information (conversation) (Stored in the database 500).
[0201]
(Sixth modification example)
When the second morpheme information including the first morpheme information cannot be searched from the second morpheme information, the topic search unit 360 collates the first morpheme information with each answer sentence, and When the answer sentence including the first morpheme information can be searched from among the first sentence, the first search means may acquire the second morpheme information associated with the searched answer sentence.
[0202]
Specifically, the topic search unit 360, to which the topic search command signal is input from the abbreviation sentence complementing unit 350, based on the first morpheme information included in the input topic search command signal, If the second morpheme information matching the first morpheme information cannot be obtained from each of the second morpheme information by comparing the morpheme information, as shown in FIG. The answer sentence associated with the second morphological information is collated.
[0203]
By this collation, when the topic search unit 360 determines that the morpheme (action or morpheme associated with the action) constituting the first morpheme information is included in the answer sentence, Search for the associated second morpheme information.
[0204]
Thus, even if the topic search unit 360 cannot search for the second morpheme information that matches the first morpheme information from each of the second morpheme information, the topic search unit 360 can search for the first morpheme information from each answer sentence. Can be identified, and the second morpheme information associated with the identified answer sentence can be searched, so that the utterance content of the user can be specified. Can be appropriately searched for the second morpheme information corresponding to the first morpheme information that constitutes.
[0205]
As a result, the topic search unit 360 can search for the optimum second morpheme information corresponding to the first morpheme information, so that the answer sentence search unit 370 determines the optimum second morpheme information searched for by the topic search unit 360. , It is possible to search for an appropriate answer sentence for the utterance content of the user.
[0206]
[program]
The contents described in the conversation control system and the conversation control method can be realized by executing a dedicated program for using a predetermined program language on a general-purpose computer such as a personal computer.
[0207]
Here, as a programming language, a hierarchical level is established by associating a topic desired by the user, the degree of emotion of the user with respect to a certain matter, or a type of statement, affirmative sentence, question sentence, repellent sentence, etc. with a morpheme according to the meaning content. In this embodiment, for example, a language for storing information in a database, such as a DKML (Discoverse Knowledge Markup Language), an XML (extensible Markup Language), and a C language developed by the inventors.
[0208]
That is, the conversation control device 1 stores data stored in each of the conversation databases 500a to 500d (storage information such as second morphological information, fixed contents, answer sentences, answer types, sets, discourse ranges, and element information), and others. Can be realized by constructing DKML (Discoverse Knowledge Markup Language), XML (eXtensible Markup Language), or the like, and executing a program for using the constructed storage information and the like.
[0209]
According to such a program according to the present embodiment, each morpheme constituting the utterance content of the user is identified, the semantic content grasped from each identified morpheme is analyzed, and the morpheme is associated with the analyzed semantic content. In general, a conversation control device, a conversation control system, and a conversation control method that have an operation effect of outputting an optimal response sentence corresponding to the utterance content of a user by outputting a pre-prepared answer sentence are generally used. It can be easily realized with a simple general-purpose computer.
[0210]
Further, the developer who develops the conversation control device 1 can hierarchically construct the second morpheme information and the like for searching for the answer sentence to the utterance content of the user using the language in the database, The conversation control device 1 can acquire an answer sentence to the utterance content from the database through a hierarchical procedure based on the utterance content of the user.
[0211]
That is, the conversation control device 1 determines and determines the hierarchy of the utterance contents of the user (for example, whether the second morpheme information stored in the database is in a higher concept or a lower concept). An appropriate answer sentence can be retrieved from each answer sentence stored in advance based on the hierarchy.
[0212]
For this reason, the conversation control device 1 does not collate the first morpheme information composed of the utterance content of the user with the “all” second morpheme information stored in advance, and each of the morpheme information belonging to a certain specific hierarchy. Since the second morpheme information and the first morpheme information need only be collated, the second morpheme information similar to the first morpheme information can be obtained in a short time.
[0213]
Further, in the communication between the communication unit 800 and the communication unit 900, data may be transmitted and received via the communication network 1000 using a protocol such as DKML. Thereby, for example, when there is no answer sentence suitable for the user's utterance content in the conversation control device 1, the conversation control device 1 is adapted to the user's utterance content through the communication network 1000 in accordance with a convention such as DKML. The searched answer sentence (written in DKML or the like) can be searched, and the searched answer sentence can be searched (see FIG. 20).
[0214]
Note that the program can be recorded on a recording medium. As shown in FIG. 23, the recording medium includes, for example, a hard disk 1100, a flexible disk 1200, a compact disk 1300, an IC chip 1400, a cassette tape 1500, and the like. According to the recording medium on which such a program is recorded, it is possible to easily store, transport, and sell the program.
[0215]
[Second embodiment]
(Basic configuration of conversation control device)
A second embodiment of the present invention will be described with reference to the drawings. As shown in FIG. 24, the conversation control device 1 according to the present embodiment has substantially the same internal structure as the conversation control device 1 according to the first embodiment, but includes a selection unit 380 (second search means; selection means). They differ in points. Further, unlike the first embodiment, the conversation control device 1 partially differs in the functions of the topic search unit 360 (first search unit) and the answer sentence search unit 370 (second search unit; answer sentence search unit). .
[0216]
Except for this difference, the structure is the same as that of the first embodiment and the modified example, and the description of the structure other than the difference will be omitted. The present embodiment naturally has various functions in the first embodiment, such as a parrot return process, a reflexive process, and the like, and also has an effect corresponding to those functions.
[0219]
In the present embodiment, the first morpheme information and topic title described later are different from the first embodiment, and each morpheme belonging to the first morpheme information and topic title is a “case” such as a subject, an object, or an action. Description will be made assuming that it is included regardless of the configuration.
[0218]
In the first embodiment, the process in which the conversation control device 1 outputs an answer sentence corresponding to the utterance content based on the utterance content from the user has been described. In the present embodiment, when the meaning of the utterance content from the user can be grasped as a plurality of meanings, the conversation control device 1 determines the utterance that is close to the topic to be developed by the conversation control device 1 from the respective meanings. The process of selecting the meaning of the content and outputting an answer sentence corresponding to the selected meaning will be described.
[0219]
Specifically, as in the first embodiment, the conversation control device 1 according to the present embodiment performs a passive answer process based on the content of the utterance from the user (for example, a process that faithfully answers the content of the question from the user). Is executed, and the conversation control apparatus 1 voluntarily performs an active answer process to guide the current topic to a specific topic. The structure of the conversation control device 1 in the present embodiment is as follows. The conversation control method and the program according to the present embodiment can be executed with the same processing content as the processing content in the conversation control device 1.
[0220]
As shown in FIG. 25, in the present embodiment, the phrase includes a topic title indicating a morpheme composed of one character, a plurality of character strings, or a combination thereof, and the topic title includes an answer sentence to the user. Corresponding. The phrase includes, for example, the utterance content or the topic title. In this embodiment, the phrase corresponds to a topic title, as shown in FIG.
[0221]
For example, as shown in FIG. 25, if the topic title 1 is (A device; IR (Inazuma Rush); teach) {Teach the IR (Inazuma Rush) of the A device}, the topic title 1 is Topic titles including some constituent morphemes (IR; teach) include topic title 2 (Company A; IR (Investor Relations); teach) {Teach IR (Investor Relations) of Company A}. Can be The above phrases, topic titles, and answer sentences are stored in the conversation database 500 (morpheme storage unit).
[0222]
The common topic titles are associated with priorities for being selected as answer sentences. In addition, it is desirable that the answer sentence includes a content for guiding the user to a specific topic or a content related to the associated topic title.
[0223]
The topic search unit 360 (first search unit) searches each phrase including the utterance content from a plurality of phrases stored in a conversation database (morpheme storage unit) in advance based on the utterance content input by the user. Is what you do. That is, in the present embodiment, the topic search unit 360, based on the first morpheme information extracted by the sentence structure analysis unit 430, selects each topic title including the first morpheme information from among the topic titles stored in advance. Search for.
[0224]
Specifically, as shown in FIG. 25, when the first morpheme information is (IR; teach), the topic search unit 360 sets the topic title 1 including the first morpheme information (IR; teach). (A device; IR; teach) and topic title 2 (Company A; IR; teach).
[0225]
The selection section 380 (selection means) selects a phrase related to a topic to be developed in advance from each phrase based on each phrase searched by the topic search section 360. That is, in the present embodiment, when a plurality of topic titles are searched by the topic search unit 360, the selection unit 380 according to the magnitude of each priority order associated with each of the plurality of searched topic titles. Select one topic title.
[0226]
Here, the fact that a plurality of topic titles including the utterance content or the first morpheme information constituting the utterance content is searched means that in the present embodiment, the utterance content is captured in a plurality of meanings. As a result, when the utterance content is captured in a plurality of meanings, that is, when a plurality of topic titles are searched, the selecting unit 380 selects a topic to be developed in advance by the conversation control device 1 from among the searched topic titles. (A topic title associated with a high priority) is selected.
[0227]
Specifically, as shown in FIG. 25, when the first morpheme information (IR; teach) is extracted by the sentence structure analysis unit 430, the topic search unit 360 transmits the first morpheme information (IR; teach). ) Are searched for. As shown in FIG. 25, the priority associated with the searched topic title 1 is 2, and the priority associated with the topic title 2 is 1, so that the selecting unit 380 selects the topic title 1 and the topic title 1. From among the topic titles 2, a topic title 2 associated with the priority order 1 is selected.
[0228]
That is, when a plurality of topic titles are searched by the topic search unit 360, the selection unit 380 specifies the highest priority from among the priorities associated with the searched topic titles. Then, a topic title associated with the specified priority is selected. In addition, the selection unit 380 may specify the lowest priority from among the above-described priorities, and select a topic title associated with the specified priority.
[0229]
The answer sentence search unit 370 (answer sentence search means) searches for an answer sentence corresponding to the phrase selected by the selection unit 380. That is, in the present embodiment, the answer sentence search unit 370 searches for an answer sentence associated with a topic title based on the topic title selected by the selection unit 380.
[0230]
More specifically, as shown in FIG. 25, for example, when a topic title 2 (Company A; IR; teach) is selected by the selection unit 380, the answer sentence search unit 370 responds to the topic title 2 The attached answer 2 (IR of company A is ~) is acquired. The output unit 600 outputs the answer sentence 2 (the IR of company A is ~) acquired by the answer sentence search unit 370.
[0231]
When the topic title searched last time is included in each topic title searched this time, the selection unit 380 selects a topic title different from the topic title searched last time from each topic title searched this time. May be elected. That is, when the topic title searched last time is included in each topic title searched this time, the selection unit 380 associates the topic title searched last time with the topic title searched last time. A priority next to the priority may be specified, and a topic title associated with the specified priority may be selected.
[0232]
More specifically, as shown in FIG. 25, for example, the selecting unit 380 determines whether the topic title 2 (Company A; IR; teach) previously searched (or selected) is the topic title 1 or the topic title searched this time. In the case of being included in No. 2, the priority 1 associated with the topic title 2 (Company A; IR; tell me) searched last (or selected) is specified. The selecting unit 380 selects the topic title 1 (A device; IR; teach) associated with the specified priority 2 next to the specified priority 1 from the topic title 1 or the topic title 2 searched this time. elect.
[0233]
(Conversation control method using conversation control device)
The conversation control method by the conversation control device 1 having the above configuration can be implemented by the following procedure. FIG. 26 is a flowchart showing the procedure of the conversation control method according to the present embodiment. The processing from (S201) to (S204) shown in FIG. 26 is the same as the processing from (S101) to (S104) in the first embodiment, and a detailed description thereof will be omitted.
[0234]
As shown in FIG. 26, based on the first morpheme information extracted by the sentence structure analysis unit 430, the topic search unit 360 selects each topic title including the first morpheme information from among the topic titles stored in advance. Is performed (S205). Specifically, as shown in FIG. 25, when the first morpheme information is (IR; teach), the topic search unit 360 sets the topic title 1 including the first morpheme information (IR; teach). (A device; IR; teach) and topic title 2 (Company A; IR; teach).
[0235]
After that, when each topic title is searched by the topic search unit 360, the selection unit 380 selects one topic title according to the magnitude of each priority order associated with each searched topic title. Is performed (S206). Specifically, as shown in FIG. 25, when the first morpheme information (IR; teach) is extracted by the sentence structure analysis unit 430, the topic search unit 360 transmits the first morpheme information (IR; teach). ) Are searched for.
[0236]
Since the priority order associated with the searched topic title 1 is 2 and the priority order associated with the topic title 2 is 1, the selecting unit 380 selects, from among the topic titles 1 and 2, The topic title 2 associated with the priority 1 is selected.
[0237]
On the other hand, when the topic title searched last time is included in each topic title searched this time, the selection unit 380 associates the topic title searched last time with the topic title searched last time. A priority next to the priority is specified, and a topic title associated with the specified priority is selected.
[0238]
Specifically, as shown in FIG. 25, for example, the selecting unit 380 includes the topic title 2 (Company A; IR; tell me) searched last time in the topic title 1 or the topic title 2 searched this time. In this case, the priority 1 associated with the previously searched topic title 2 (Company A; IR; tell me) is specified. The selecting unit 380 selects the topic title 1 (A device; IR; teach) associated with the specified priority 2 next to the specified priority 1 from the topic title 1 or the topic title 2 searched this time. elect.
[0239]
Next, the answer sentence search unit 370 performs a step of searching for an answer sentence associated with the topic title based on the topic title selected by the selection unit 380 (S207). More specifically, as shown in FIG. 25, for example, when a topic title 2 (Company A; IR; teach) is selected by the selection unit 380, the answer sentence search unit 370 responds to the topic title 2 The attached answer 2 (IR of company A is ~) is acquired. The output unit 600 outputs the answer sentence 2 (the IR of company A is ~) acquired by the answer sentence search unit 370.
[0240]
(Operation and effect of the conversation control device and the conversation control method)
According to the invention having the above configuration, as each topic title including the first morpheme information (IR; teach) extracted by the sentence structure analysis unit 430, the topic title 1 (A device; IR (inazuma rush); ) And topic title 2 (Company A; IR (Investor Relations); tell me), the selection unit 380 selects a higher priority from among the retrieved topic titles 1 and 2 The attached topic title 2 can be selected (see FIG. 25).
[0241]
Here, the fact that a plurality of phrases each including input information is searched, or that a plurality of second morphological information including first morphological information constituting the input information are searched for, means that the input information has a plurality of meanings. Means that. Therefore, when the utterance content is captured in a plurality of meanings, that is, when a plurality of topic titles are searched, the selection unit 380 wants to develop in advance the conversation control device 1 from among the topic titles as described above. A topic title related to the topic (a topic title associated with a high priority) can be selected. As a result, the answer sentence search unit 370 can output an answer sentence associated with the topic title (for example, a sentence for guiding to a specific topic) based on the topic title selected by the selection unit 380. .
[0242]
Also, the conversation control device 1 does not output a passive answer sentence based on the content of the utterance from the user (for example, faithfully answer the content of the question from the user). It is possible to output an active answer sentence that guides the user from a current topic to a specific topic.
[0243]
Further, since the answer sentence includes content for guiding the user to a specific topic, the answer sentence search unit 370 outputs the answer sentence via the output unit 600 to expand the current topic in advance. You can more accurately guide you to the topic you want to make. That is, the conversation control device 1 can internally take the initiative on the current topic and guide the current topic to a specific story more accurately.
[0244]
Furthermore, the answer sentence search unit 370 searches for a corresponding answer sentence using morphemes excluding Japanese-specific particles (te, ni, wo, etc.) included in the input information. The answer sentence created in advance can be acquired more accurately based on the semantic content that can be grasped from the user. That is, the answer sentence search unit 370 can search for an answer sentence more suitable for the input information than simply using the entire input information as a keyword and searching for an answer sentence associated with the keyword.
[0245]
In addition, since the topic search unit 360 searches for a topic title including the first morphological information, there is no need to search for a topic title that completely matches input information from the user. As a result, the developer who develops the conversation control device 1 does not need to previously store a huge response sentence corresponding to the input information expected to be input by the user, and can reduce the capacity of the storage unit. it can.
[0246]
(Example of change)
Note that the present invention is not limited to the above embodiment, and the following changes can be made. When each topic title searched this time by the topic search unit 360 is searched for more than a predetermined reference number of times, the selection unit 380 selects a topic title different from the previously selected topic title in each of the currently searched topic titles. It may be selected from among topic titles.
[0247]
Specifically, the selection unit 380 determines that the topic title 1 (A device; IR; tell me) selected last time is included in the topic title 1 or topic title 2 searched this time, and the topic title searched this time, for example. 1 or a topic title 2 is specified as being searched for more than a predetermined reference number of times (for example, 5 times).
[0248]
When the specified number of times exceeds a predetermined reference number (for example, five times), the selection unit 380 determines the priority order associated with the previously selected topic title 1 (A device; IR; teach). 1 is specified. The selecting unit 380 selects a topic title 2 (Company A; IR; tell me) associated with the specified priority 2 next to the specified priority 1 from the topic title 1 or the topic title 2 searched this time. I do.
[0249]
On the other hand, if the counted number does not exceed the reference number (for example, 5 times), the selecting unit 380 replaces the previously searched topic title 1 (A device; IR; teach) with another topic title 2, Used in preference to topic title 3. The answer sentence search unit 370 searches for an answer sentence associated with the topic title based on the topic title selected by the selection unit 380. The answer sentence search unit 370 outputs the answer sentence obtained via the management unit 310 to the output unit 600.
[0250]
Thus, if the topic title selected last time is included in each topic title searched this time by the topic search unit 360, and each topic title searched this time is searched for more than a predetermined reference number, the selection unit 380 can select a topic title different from the topic title selected last time from each topic title searched this time.
[0251]
That is, when the same utterance content is repeatedly uttered by the user, the selection unit 380 can sequentially change the topic title in accordance with the number of times the utterance is repeated. As a result, the conversation control device 1 can output an answer sentence for guiding a current topic to a specific topic in accordance with the number of repeated utterance contents. Also, when the same utterance content is repeatedly uttered, the conversation control device 1 guides from the current topic to another topic, so that the user can feel as if talking with a human. .
[0252]
Note that the present invention is not limited to a device or the like (conversation robot or the like) for the purpose of having a conversation, and this function can be used in a game machine or the like.
[0253]
【The invention's effect】
As described above, according to the present invention, when the content of a user's utterance is captured in a plurality of meanings, an answer sentence related to a specific topic to be developed in advance is output, so that the current topic can be specified. Can guide you to the topic.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a schematic configuration of a conversation control system according to a first embodiment.
FIG. 2 is a block diagram illustrating an internal structure of a conversation control unit and a sentence analysis unit according to the first embodiment.
FIG. 3 is a diagram illustrating the content of each morpheme extracted by a morpheme extraction unit according to the first embodiment.
FIG. 4 is a diagram showing the contents of each phrase extracted by a phrase analysis unit in the first embodiment.
FIG. 5 is a diagram illustrating contents of “case” specified by a sentence structure analysis unit according to the first embodiment.
FIG. 6 is a diagram illustrating “utterance sentence type” specified by an utterance type determination unit according to the first embodiment.
FIG. 7 is a diagram showing the contents of each dictionary stored in the utterance type database in the first embodiment.
FIG. 8 is a diagram showing contents of a hierarchical structure constructed inside a conversation database in the first embodiment.
FIG. 9 is a diagram showing a detailed relationship of a hierarchical structure constructed inside a conversation database in the first embodiment.
FIG. 10 is a diagram showing the contents of a “topic title” constructed inside a conversation database in the first embodiment.
FIG. 11 is a diagram showing the contents of “type of answer sentence” associated with “topic title” constructed inside the conversation database in the first embodiment.
FIG. 12 is a diagram showing contents of “topic title” and “answer sentence” belonging to a “discussion range” constructed inside a conversation database in the first embodiment.
FIG. 13 is a diagram illustrating contents of reflection element information stored in a reflection element database according to the first embodiment.
FIG. 14 is a diagram showing parrot return elements stored in a parrot return element database and the contents of morphemes of the parrot return elements in the first embodiment.
FIG. 15 is a flowchart illustrating a procedure of a conversation control method according to the first embodiment.
FIG. 16 is a diagram showing utterance contents arranged by a morpheme extraction unit in the first modification.
FIG. 17 is a diagram illustrating an internal configuration of a topic search unit according to a second modification.
FIG. 18 is a diagram illustrating a manner in which the ratio calculation unit in the second modification checks each morpheme belonging to “case configuration” and each “topic title” for each “topic title”.
FIG. 19 is a diagram illustrating a manner in which the ratio calculation unit according to the second modified example collates each morpheme belonging to “each component” with each morpheme belonging to “topic title” for each “case”.
FIG. 20 is a diagram illustrating a schematic configuration of a conversation control system according to a third modification.
FIG. 21 is a diagram illustrating an internal configuration of a topic search unit in a fifth modification.
FIG. 22 is a diagram illustrating a manner in which a topic search unit in the sixth modification example compares first morpheme information with a topic title or an answer sentence.
FIG. 23 is a diagram illustrating a recording medium that stores a program according to the first embodiment.
FIG. 24 is a block diagram illustrating an internal configuration of a conversation control device according to a second embodiment.
FIG. 25 is a diagram showing contents of “topic title” and “answer sentence” constructed inside a conversation database in the second embodiment.
FIG. 26 is a diagram showing a processing procedure of a conversation control method according to the second embodiment.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Conversation control device, 2 ... Server, 100 ... Input part, 200 ... Speech recognition part, 300 ... Conversation control part, 310 ... Management part, 320 ... Reflective judgment part, 330 ... Parrot return judgment part, 340 ... Discourse range determination Section, 350: abbreviated sentence complementing section, 360: topic search section, 361: ratio calculating section, 362: selecting section, 363: deleting section, 364: discourse adding section, 370: answer sentence searching section, 380: selecting section, 400 ... Sentence analysis unit, 410 ... Morphological extraction unit, 420 ... Phrase analysis unit, 430 ... Sentence structure analysis unit, 440 ... Utterance type determination unit, 450 ... Morphological database, 460 ... Utterance type database, 500 ... Conversation database, 600 ... Output Unit, 700: voice recognition dictionary storage unit, 800: communication unit, 801: reflection element database, 802: parrot return element database, 900: communication unit, 1000: communication Network, 1100 ... hard disk, 1200 ... flexible disk, 1300 ... compact disc, 1400 ... IC chip, 1500 ... cassette tape

Claims (5)

A conversation control device that searches for an answer sentence corresponding to the input information based on input information including utterance content input from a user,
Based on the input information input from the user, from among a plurality of phrases stored in advance, a first search means for searching for each phrase including the input information,
A second step of selecting a phrase related to a topic to be developed in advance from each of the phrases based on the respective phrases searched by the first search unit, and searching for the answer sentence corresponding to the selected phrase; A conversation control device comprising a search unit. The conversation control device according to claim 1, wherein
A morpheme extracting unit configured to extract at least one morpheme constituting a minimum unit of the character string as first morpheme information based on the character string included in the input information;
The phrase includes second morpheme information indicating a morpheme composed of one character, a plurality of character strings, or a combination thereof, and the plurality of second morpheme information is selected as the answer sentence and the answer sentence. And a morphological storage means for storing the respective second morphological information and the respective answer sentences in advance,
The first retrieval unit is configured to select the second morpheme including the first morpheme information from among the second morpheme information stored in advance based on the first morpheme information extracted by the morpheme extraction unit. Search for information,
The second search means, when the first search means has searched for each of the second morpheme information, according to the magnitude of each of the priorities associated with each of the searched second morpheme information Selecting means for selecting one piece of the second morphological information from among the pieces of the second morphological information;
A conversation control device, comprising: an answer sentence search unit that searches for the answer sentence associated with the second morpheme information based on the second morpheme information selected by the selection unit. The conversation control device according to claim 2, wherein
The selecting means, when each of the second morphological information searched this time by the first searching means is searched for more than a predetermined reference number of times, the second morphological information different from the second morphological information selected last time. A conversation control device, wherein dimorphological information is selected from each of the second morphological information searched this time. The conversation control device according to claim 1, wherein:
The conversation control device, wherein the answer sentence includes content for guiding the user to a specific topic. A conversation control method for searching for an answer sentence corresponding to the input information based on input information including utterance content input from a user,
A first step of searching for each phrase including the input information from a plurality of phrases stored in advance based on the input information input from the user;
A second step of selecting a phrase related to a topic to be developed in advance from the respective phrases based on the respective phrases searched in the first step, and searching for the answer sentence corresponding to the selected phrase; And a conversation control method.

Images