JPH01161479A - Natural language interactive device - Google Patents

Abstract

PURPOSE:To form an interactive sentence using a proper honorific expression based on respective social relation by storing a belonging relation to the remote side of an interaction and a person appearing in the interaction and storing information for using an expression from these relations. CONSTITUTION:At the time of forming the sentence, an expression decision device 4 extracts a group to which the device itself and the remote side of the interaction belong from an subject belonging storage device 5, extracts a group to which the person appearing in the sentence belongs from a general belonging storage device 6, further extracts a honorific expression and a polite representation to be used to the person having the assignment relation from an expression storage device 7 and forms a natural language sentence based on the expression. Thereby, the natural language sentence using the proper honorific expression and the polite expression can be formed to the remote side of the interaction and the appearing person.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention is directed to the use of a natural language dialogue, in which the conversation partner and the person present in the conversation, the group and device to which they belong, and the group and apparatus to which the dialogue partner belongs. ′ in the dialogue and its characters according to their affiliation with the group they are in! ! ! This invention relates to a natural language dialogue device that generates polite honorific expressions and polite expressions.

Conventional technology Conventional devices that generate sentences using honorific expressions or polite expressions use honorific expressions unconditionally and uniformly for the user or a specific person. Even if the incident was within one's own family, it could not be handled based on the principle of using honorific language, such as not using honorific language in conversations with outsiders.

Problems that the invention aims to solve In general, in conversation, expressions of respect or politeness may or may not be used toward the same person depending on the social affiliation relationship between that person and the speaker and the social relationship with the listener. or However, with conventional devices, it was necessary to specify whether to use expressions of respect, respect, or politeness for the characters appearing.
It was not possible to change the expression appropriately according to the relationship between the characters and the characters.

Means for Solving the Problems The natural language dialogue device of the present invention has a subject attribution storage device that stores the group to which the device belongs and the group to which the person with whom the device is interacting belongs, and a general group that stores the group to which the person appearing in the sentence belongs. Attribution storage device, an expression storage device that stores rules for what kind of honorific expressions and polite expressions should be used for a person with a specific attribution relationship, and an expression that determines the expression of a natural language sentence to be generated based on the rules. It has a determining device.

When generating an action sentence, the expression determining device extracts the group to which the device itself and the interacting partner belong from the subject attribution storage device,
The group to which the person appearing in the sentence belongs is extracted from the general attribution storage device, the respectful expressions and polite expressions that should be used for the person who has such an attribution relationship are extracted from the expression storage device, and the table thereof is extracted. By generating a natural language sentence based on TM, it becomes possible to generate a natural language sentence using appropriate expressions of respect and politeness for the person with whom the user interacts and the characters appearing.

Embodiment In the embodiment shown in FIG. 1, a person with whom one is interacting manually writes a natural language sentence using a human-powered device 1, and the human-powered sentence analysis device 2 analyzes the natural language input sentence. Based on the results of the analysis, the response sentence generation device 3 generates a response sentence expression for the input sentence, and the expression determination device 4 generates a response sentence expression for the input sentence.
extracts the affiliation relationship between the device itself and the person it interacts with from the subject attribution storage device 5, roughly extracts the affiliation relationship of the person appearing in the response sentence from the general attribution storage device 6, and expresses the expressions for the characters and the conversation partner. Based on the expression extracted from the expression storage device 7, the expression of the response sentence is converted into an actual response sentence and output to the output device 8.

FIG. 2 specifically shows the contents of the subject attribution storage device 5, and shows that the device belongs to Section F of Department D of Company B;
This indicates that the conversation partner belongs to Family A and Section 0 of Department E of Company C.

Figure 3 specifically shows the contents of the general attribution storage device 6, and it says, ``A person named J belongs to Section F, Department, Department of Company B, and ``A person named A Takahashi belongs to Section Z, Department Y, Company X.'' Indicates that the person belongs to.

The actual operation of the apparatus in the embodiment shown in FIGS. 1 to 3 will be explained below with reference to the following example.

Dialogue partner: "Mr. Masume, would you like to come?" Device: "Mr. Masume is currently away from his seat." The sentence input from the human-powered device 1 is analyzed by the human-powered sentence analysis device 2, and the result is " It turns out that the existence of a person named square j is being asked.Based on this, the response sentence generation device 3 decides to state that ``square'' is currently absent. The expression determining device 4 extracts the affiliation relationship between the device itself and the dialogue partner from the subject attribution storage device 5, and furthermore, the expression determination device 4 extracts the affiliation relationship of “square” from the general attribution storage device 6. For example, the expression storage device 7 has a rule such as ``If the other party in the conversation belongs to a company different from the device itself, polite expression El should not be used for a person who belongs to the same company as the device itself.'' The expression determining device refers to these rules and decides to use respectful expressions for the dialogue partner, and not to use polite expressions for square j, etc. Based on this, the expression determination device uses the above example. Generate a response sentence like .

The embodiment shown in FIG. 4 is an example in which the stored contents of the subject attribution storage device and the general attribution storage device are rewritten based on the analysis results of the input sentence analysis section. When the device interacts with an unspecified party, it is impossible to store in advance the affiliation information of the party to be interacted with in the subject attribution storage device 5. In this case, it is necessary to identify the conversation partner based on an input sentence from the other party, and further determine the affiliation of the other party. for example,
At the beginning of the dialogue, ask and answer questions such as the following:

Device: “Who are you?” Interaction partner: “My name is Takahashi.” As a result of human-powered sentence analysis device 2 analyzing the input sentence from the other party,
It turns out that the conversation partners are r Takahashi and fi. The human sentence analysis device 2 sends this information to the storage management device 9. The storage management device 9 extracts the affiliation relationship of the person named Takahashi J from the general attribution storage device 6 and registers this as the affiliation relationship information regarding the dialogue partner in the subject attribution storage device @5.
Afterwards, you can use appropriate honorific expressions in your responses to the other person.

Further, it is possible that information regarding the person who appeared in the dialogue is not stored in the general attribution storage device 6. In this case as well, it is conceivable to add appropriate information to the general attribution storage device based on the result of analyzing the input sentence.

For example:

Dialogue partner = ``Isn't Drawer from our company listening to you?'' Device: ``Mr. Drawer is not here.'' Human-powered sentence analysis device 2 asks the other person. The input sentence is analyzed, and information is obtained that the person named rquotej belongs to the same company as the person with whom the conversation is taking place. The human sentence analysis device sends this information to the storage management device 9, and the storage management device 9 says, ``If the information regarding drawer j is not stored in the general attribution storage device 6, add that information to the general attribution storage device. This makes it possible to use appropriate honorific expressions when that person appears in subsequent dialogue.

Note that the example shown in FIG. 4 above is similar to the example shown in FIG. 1 with regard to the movement of the expression determining device 4.

Effects of the invention By using the above device, the device itself, the conversation partner,
By memorizing the affiliations of people appearing in dialogue, and further memorizing information on what expressions should be used based on those relationships, it is possible to use appropriate honorific expressions based on each social relationship. It becomes possible to generate dialogue sentences that

[Brief explanation of the drawing]

FIG. 1 is a conceptual diagram of a natural language dialogue device according to an embodiment of the present invention, FIG. 2 is a conceptual diagram of a subject attribution storage device in the same natural language dialogue device, and FIG. 3 is a conceptual diagram of a general attribution storage device in the same device. FIG. 4 is a conceptual diagram of another embodiment of the present invention. l:...Human power device, 2...Human power sentence analysis device; 3...
Response sentence generation device, 4... Expression determining device, 5... Subject attribution storage device, 6... General attribution storage device, 7... Expression storage device, 8... Output device, 9... Storage management device.

Claims (1)

[Claims] In a device that performs dialogue using natural language sentences, a dialogue subject attribution memory device that stores the social group to which the device belongs and the social group to which the person with whom the device interacts belongs, and the affiliation of the person appearing in the generated sentence. A general attribution memory device that remembers social groups, an expression memory device that stores rules about what kind of honorific expressions and polite expressions should be used for people who belong to each group, and nature generated based on those rules. A natural language device characterized by having an expression determining device that determines the expression of a natural language sentence, and generating appropriate polite expressions and polite expressions for the person with whom it interacts and the person appearing in the sentence when generating a natural language sentence. Language dialogue device.