JPH08235205A - Interactive sentence preparation method and natural language intractive device - Google Patents

Abstract

PURPOSE: To provide an interactive sentence preparation method and a natural language interactive device for responding with appropriate expression by considering the personal relation of a user and a retrieval object person or the user and a computer. CONSTITUTION: In this interactive sentence preparation method for preparing the response sentence of a retrieved result in the case of retrieving information by interaction, the personal relation of a retrieval instructing person who instructs retrieval and the retrieval object person to be the object of the retrieval is extracted from the input of a retrieval instruction (301 and 302) and the expression of the response sentence of the retrieved result is changed and outputted corresponding to the extracted personal relation (303 and 106.) The personal relation is extracted based on the expression including the expression of names and the expression for indicating a managerial position name and intimacy in an input sentence. Or, the personal relation is extracted based on the number of times of inputting the retrieval instruction.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dialog sentence creating method for creating a dialog sentence when information is retrieved through a dialog in a natural language and a natural language dialog device for realizing the dialog sentence creating method.

[0002]

2. Description of the Related Art In recent years, with the expansion of computer networks and the development of personal information terminals, it has become possible to retrieve various personal information such as contact information and schedule management. In general, when such personal information retrieval is performed by natural language dialogue through a natural language dialogue device, the system usually generates a response sentence based on a fixed expression.

[0003]

Therefore, in the above-mentioned conventional apparatus, when a polite expression is required, the user may be disrespected by using the distracting expression, or the expression may be too polite. There are drawbacks such as making it happen. An object of the present invention is to eliminate the above-mentioned conventional drawbacks and provide a dialogue sentence creating method and a natural language dialogue device for responding with appropriate expressions in consideration of interpersonal relationships between a user and a search target person or a user and a computer. To do.

[0004]

In order to solve the above-mentioned problems, a dialog sentence creating method of the present invention is a dialog sentence creating method for creating a response sentence of a search result when information is searched by a dialog. From the input of the search instruction, the interpersonal relationship between the search instructor who instructed the search and the search target person who is the search target is extracted, and the expression of the response sentence of the search result is changed corresponding to the extracted interpersonal relationship. It is characterized by outputting as.

Here, the extraction of the interpersonal relationship is performed based on an expression of a name in an input sentence or an expression including a position name and an expression of friendship. Also, the extraction of the interpersonal relationship is
First, storing correspondence between expressions in the input sentence and interpersonal relationships
The storage means is referred to. Further, the output of the response sentence is performed by referring to the second storage means for storing the correspondence between the name, the post name, the interpersonal relationship and the expression of the response sentence.
Further, the extraction of the interpersonal relationship and the output of the response sentence are performed with reference to the third storage means for storing the correspondence between the expression in the input sentence and the expression of the response sentence. The extraction of the interpersonal relationship is performed based on the number of times the search instruction is input.

Further, the natural language interactive apparatus of the present invention is a natural language interactive apparatus for performing information retrieval by dialogue, in which a search instructor who has instructed a search and a search target person who is a search target are input from a search instruction input. It is characterized by comprising interpersonal relation extracting means for extracting interpersonal relations, and response sentence output means for changing the expression of the response sentence as the retrieval result and outputting it in correspondence with the extracted interpersonal relations.

Here, the interpersonal relationship extracting means includes a first storage means for storing the interpersonal relationship and the expression including a name expression or a position name and an expression expressing familiarity in the input sentence in association with each other. The interpersonal relationship is extracted based on the expression of the input sentence. Further, the response sentence output means includes a second storage means for storing the name, the post name and the extracted interpersonal relationship and the expression of the response sentence in association with each other, and the name, the post name and the extracted interpersonal relationship. Based on, the expression of the response sentence is changed and output. Further, the interpersonal relationship extracting means includes a comparing means for comparing the number of times of inputting the search instruction with a threshold value, and extracts the interpersonal relationship based on the number of times of input.

Further, the natural language interactive apparatus of the present invention is a natural language interactive apparatus for performing information retrieval by dialogue, and based on the interpersonal relationship between the search instructor who has instructed the search and the search target person who is the search target, the input is performed. Third storage means for storing the correspondence between the expression in the sentence and the expression of the response sentence and the input sentence of the search instruction are referred to the third storage means, and the expression of the response sentence of the search result is changed and output. And a response sentence creating means.

[0009]

Embodiments of the present invention will be described below in detail with reference to the drawings. <Example of Configuration of Natural Language Dialogue Apparatus> FIG. 1 is a diagram showing an example of a block configuration of the natural language dialogue apparatus of the present embodiment. In the figure, 101 is a sentence input unit for inputting an input sentence in natural language, and 102 is a display unit for outputting an output sentence, a search result, information indicating a flow of dialogue, and the like. 103
Is a sentence analysis unit that analyzes the input sentence input by the sentence input unit 101. An information search unit 104 includes a database 104a and a search unit 104b, creates a search instruction from the analysis result analyzed by the sentence analysis unit 103, and searches the information stored in the database.

Reference numeral 105 extracts input interpersonal information (name, familiar expression) from the analysis result analyzed by the sentence analysis unit 103, determines interpersonal relationships among users / systems / search target persons, and outputs them as output sentences. It is an interpersonal relationship determination unit that generates output interpersonal information (name, tone) used. 106 is the information retrieval unit 1
The sentence generation unit generates an output sentence based on the search result retrieved in 04 and the interpersonal information determined by the interpersonal relationship determination unit 105. Reference numeral 107 denotes the analysis result analyzed by the sentence analysis unit 103, the information search unit 104 and the interpersonal relationship determination unit 1.
05 is a dialogue management unit that gives a sentence generation instruction and responds to a dialogue.

FIG. 2 is a diagram showing an example of the hardware configuration of the natural language dialog device of this embodiment. In the figure, 201 is a CPU for arithmetic / control that controls the entire apparatus, 202 is a ROM that stores fixed values and control programs, such as a dialogue management program 202a, a minute analysis program 202b, an interpersonal relationship determination program 202c, and a sentence creation program. 202
d, a search program 202e, etc. are stored. These programs are only the main ones of this embodiment, and the system program, the input / output control program, etc. are not shown. CPU201 according to these programs
1 is realized by the control of FIG.

Reference numeral 203 denotes a RAM for auxiliary storage or temporary data storage, for example, a search key storage unit 203a for storing a search key extracted from an input sentence, for extracting an interpersonal relationship (parent / sparse) from the input sentence. Friendship determination table 20
3b, an output expression table 203c that determines the expression of an output sentence from interpersonal relationships, input nomenclature, etc., an input sentence storage unit 203d that stores the input sentence, and an output sentence storage unit 203 that stores the output sentence.
Including e. Note that the tables 203b and 203d may be stored in the ROM 202 if they are fixed. A large-capacity storage unit 204 such as a hard disk or a floppy disk stores a database to be searched in this embodiment. Further, the programs on the ROM 202 are stored in the large-capacity storage unit 204 and stored in the RAM 203.
May be loaded and executed on.

Reference numeral 205 denotes an input interface, which is a keyboard 101 for inputting operation instructions and data of this apparatus.
a, a voice input unit 101b for performing voice input, and the like are connected. In addition to the above key input or voice input, the input sentence of this example may be an image input from a scanner or online input via a communication line. An output interface 206 is a display unit 10 that displays operation instructions and search results.
2, a voice output unit 207 that outputs a voice, a printer 208 that outputs a hard copy, and the like are connected. The output may be an online output via a communication line.

(Example of Configuration of Interpersonal Relationship Determining Section) FIG. 3 is a diagram showing a detailed configuration of the interpersonal relationship determining section 105. The same elements as those in FIG. 1 have the same reference numerals. However,
Shows the configuration of the device in the common bus format, while FIG. 3 is expressed to clarify the flow of data between the elements of the device.

Reference numeral 301 is an input interpersonal information extraction unit for extracting interpersonal information from the analysis result analyzed by the sentence analysis unit 103 to determine interpersonal relationships, and 302 is an interpersonal relationship holding unit for holding interpersonal relationships. . 303 is an interpersonal relationship holding unit 30.
2 is an interpersonal information determination unit that determines interpersonal information used in an output sentence from the interpersonal relationship held in 2. The input interpersonal information extraction unit 301 extracts the following interpersonal information, determines interpersonal relationships (see FIG. 4), and holds the interpersonal relationship holding unit 303.

-Friendly expression-The colloquial expression for a close partner such as "Tell me / What" is "close", and the expression such as "Show (imperative) / Tell me (polite)" is "Disjoint" Hold as. It should be noted that the initial value of the familiar expression is “distant”, and the previous value is used when the familiar expression does not appear in the sentence analysis result.

-Name-A suffix that is suffixed to a job title such as "President / Manager / Teacher" and a person name such as "Sama / Mr / Mr / Kim" is extracted and the character string is retained. The output interpersonal information determination unit 302 determines the name and tone used in the output sentence by using the target table shown in FIG. 5 below.

(Familiarity Judgment Table) FIG. 4 is a diagram showing an example of the configuration of the friendship judgment table 203b. In this table, colloquial expressions for close friends such as "Tell me / What" is "close", and expressions such as "Show (imperative) / Tell me (polite)" are judged as "distant" Correspondence to do is retained.

(Output Expression Table) FIG. 5 is a diagram showing an example of the structure of the output expression table 203c. In this table, "closeness" determined by the above-mentioned familiarity determination table
The expression of the output sentence (name or isomasu / is available) from the "alienation" and post titles such as "President / Manager / Teacher" and suffixes such as "Sama / Mr / San / Kimi" Correspondence for selecting (including tones such as masquerade) is retained.
Here, (ni1) indicates that nothing is given as a name (that is, a kind of humble expression).

<Operation Example of Natural Language Dialogue Apparatus> Next, an operation example of the natural language dialogue apparatus configured as described above will be described with reference to the flowchart of FIG. First, step S
At 201, when an input sentence is input from the input unit 101, the input sentence is sent to the sentence analysis unit 103. In step S202, morphological analysis and syntactic analysis are performed on the input sentence. Then, in step S203, the information search unit 104
In, a search instruction is created from the analysis result analyzed by the sentence analysis unit 103, the information stored in the database is searched, and the process proceeds to step S204.

Here, the information search unit 104 creates a search instruction from the management information of the database and the syntax information analyzed by the sentence analysis unit 103 (this is included in the analysis result). For example, in the input sentence "I want to know Mr. Suzuki's extension in the general affairs department," the expression "I want to know" determines that it is a search request, and the expression "extension" refers to database management information. Select "Extension table". Also, as the search condition, "Suzuki" from the suffix "san" of the personal name of "Mr. Suzuki" is determined as the condition of the name item, and the suffix "department" of the department belonging to "general affairs department" is changed to "general affairs department".
Is determined as the condition of the belonging item. (If such syntax information does not exist, the item name is queried back.) Further, as a search symmetrical item, a position is added as information used by the interpersonal relationship extracting unit 301.

Therefore, assuming that the database is RDB, the following SQL is generated from the above input. select
(Extension, position) from (extension table) where (name = Suzuki,
(Affiliation = general affairs department) In step S204, the interpersonal relationship determination unit 105 extracts interpersonal information (name, familiar expression) from the analysis result analyzed by the sentence analysis unit 103, and based on the above-mentioned familiarity determination table, the user / The interpersonal relationship between the system / search target person is determined, and in step S205, interpersonal information (name, tone) used in the output sentence is generated based on the output expression table. As described below, steps S204 and S205 can be one step.
In step S206, the sentence generation unit 105 generates an output sentence based on the search information retrieved by the information retrieval unit 104 and the interpersonal information determined by the interpersonal relationship determination unit 105. Then, in step S207, the output sentence, the search result, the information indicating the flow of the dialogue, and the like are output to the display unit 102 (or another medium), and the process returns to step S201.

<Specific Example of Output Sentence Generation> As one specific example of output sentence generation, an example of designation and tone determination will be shown below. (1) "Tell me the extension of Mr. Tanaka of the Human Resources Department." → (Familiar expression of input = familiarity, name = you) → (Name of output = Mr,
(Tone = masuda key) → "Tanaka's extension is xxx." (2) "Show Tanaka's extension in the personnel department." = (Ni
1), word tone = there is tone) → "Tanaka's extension is xxx
It is. (3) “What is the extension number of Director Sato of the Planning Department?” →
(Familiar expression of input = closeness, name = manager) → (Name of output = manager, tone = Damasuki) → "Extension of Sato is xx
It is xx. (4) “Please tell me the extension of Director Sato of the Planning Department.” →
(Familiar expression of input = alienation, name = you) → (Name of output =
Director, word style = Yes tone) → "The extension of Director Sato is xx
It is xx. "

[0024]

[Other Examples]

1. In the present embodiment, the friendship determination table and the output expression table are separately provided, but this table may be a single table. In this case, the result of the friendship determination does not appear on the surface and the expression of the input sentence is used. The output representation is generated directly. 2. In the above embodiment, the friendly expressions are classified into binary values, but 3
You may classify more than the value. In this case, the word to be used in the output includes "da key", "da key", and the like.

3. Further, in the above embodiment, the target tables of FIGS. 4 and 5 are treated as fixed ones, but other examples such as
The user may change the values in the table. 4. Further, in the above embodiment, the sparse expression is determined only from the information that appears in the input sentence, but in other cases, for example, it is determined by the number of times the system is accessed, and if the access is performed a certain number of times or more, "closeness" is assumed. I don't mind. By adding this example to the above embodiment, it is possible to realize a dialogue using an appropriate expression that takes into consideration not only the user and the search target person but also the interpersonal relationship between the user and the computer. Further, in the above embodiment, the name is determined based on the post name and suffix stored in the normal natural language processing dictionary, but it may be determined using the personal name dictionary or the user dictionary in which the nickname is registered. good.

6. Further, in the above embodiment, the friendly expression and the name are initialized for each dialogue, but they may be saved by using an external storage device, loaded appropriately, and switched. In that case, save / save for each user and search target
Needless to say to load. Although the present embodiment has been described based on a simple concrete example, the present invention discloses a technical idea of generating an output sentence corresponding to an interpersonal relationship with a user (search instructor). , It also includes more precise analysis of input sentences and generation of output sentences. Further, the present invention may be applied to a system including a plurality of devices or an apparatus including one device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0027]

As described above, according to the present invention,
It is possible to provide a dialogue sentence creating method and its natural language dialogue device that respond with appropriate expressions in consideration of the interpersonal relationship between the user and the search target person or the user and the computer. That is, the interpersonal relationship is determined from the analyzed input sentence of the search instruction or the number of times of the search instruction, and at the time of sentence generation, the output sentence is generated and output based on the information retrieval result and the interpersonal relation. There is an effect that it is possible to realize a dialogue using an appropriate expression in consideration of interpersonal relationships between users or between a user and a computer.

[Brief description of drawings]

FIG. 1 is a block diagram of a natural language dialog device according to an embodiment.

FIG. 2 is a hardware configuration diagram of the natural language dialog device according to the present embodiment.

FIG. 3 is a diagram illustrating a detailed configuration example of an interpersonal relationship determination unit.

FIG. 4 is a diagram showing an example of a friendship determination table.

FIG. 5 is a diagram showing an example of an output expression table.

FIG. 6 is a flowchart showing an example of a procedure of interactive document processing according to this embodiment.

[Explanation of symbols]

 101 sentence input unit 102 display unit 103 information retrieval unit 104 sentence analysis unit 105 interpersonal relationship determination unit 106 sentence generation unit 107 dialogue management unit

Claims (11)

[Claims] 1. A dialog sentence creating method for creating a response sentence as a search result when information is searched by dialogue, comprising: a search instructor who has instructed to search and a search target to be searched after inputting a search instruction. A method for creating an interactive sentence, comprising extracting an interpersonal relationship with a person, and changing the expression of a response sentence as a search result in accordance with the extracted interpersonal relationship and outputting the response sentence. 2. The dialog sentence creating method according to claim 1, wherein the interpersonal relationship is extracted based on an expression of a name in the input sentence or an expression including a position name and an expression of friendship. 3. The dialogue according to claim 1, wherein the interpersonal relationship is extracted by referring to a first storage unit that stores the correspondence between the expression in the input sentence and the interpersonal relationship. How to create a sentence. 4. The output of the response sentence is performed by referring to a second storage unit that stores the correspondence between the name, the job title, and the interpersonal relationship and the expression of the response sentence.
How to create the interactive sentence described. 5. The extraction of the interpersonal relationship and the output of the response sentence are performed by referring to a third storage means for storing the correspondence between the expression in the input sentence and the expression in the response sentence. The method for creating a dialogue sentence according to item 2. 6. The dialogue sentence creating method according to claim 1, wherein the interpersonal relationship is extracted based on the number of times a search instruction is input. 7. A natural language dialog device for interactively searching for information, wherein an interpersonal relationship extracting means for extracting an interpersonal relationship between a search instructor who has instructed a search and a search target person who is a search target by inputting a search instruction. And a response sentence output means for changing the expression of a response sentence as a search result and outputting the response sentence according to the extracted interpersonal relationship. 8. The interpersonal relationship extracting means includes a first storage means for storing the interpersonal relationship in association with an expression including a name expression or a position name and an expression representing familiarity in an input sentence, 8. The natural language dialogue apparatus according to claim 7, wherein the interpersonal relationship is extracted based on the expression of a sentence. 9. The response sentence output means comprises a second storage means for storing the name, the post name, the extracted interpersonal relationship and the expression of the response sentence in association with each other, and the name, the post name and the extracted 9. The natural language dialogue apparatus according to claim 8, wherein the response sentence expression is changed and output based on the interpersonal relationship. 10. The interpersonal relationship extracting means includes a comparing means for comparing the number of times of inputting a search instruction with a threshold value, and the interpersonal relationship is extracted based on the number of times of inputting. Natural language dialogue device. 11. A natural language dialog device for interactively searching for information, and an expression in an input sentence and an expression in a response sentence based on an interpersonal relationship between a search instructor who instructed a search and a search target person who is a search target. And a response sentence creating means for changing the expression of the response sentence of the search result from the input sentence of the search instruction and changing the expression of the response sentence of the search result, and outputting the response sentence. Characteristic natural language dialogue device.