KR100768731B1 - A VoiceXML Dialogue apparatus based on Speech Act for Controlling Dialogue Flow and method of the same - Google Patents

Abstract

The present invention relates to the field of voice conversation interface, and more particularly, to a speech act-based VoiceXML (Voice Extensible Markup Language) conversation apparatus and method for controlling conversation flow. And a VoiceXML module for controlling the dialogue flow of the dialogue manager based on dialogue act information extracted from the dialogue contents through DDML (Dialogue Description Markup Language).

DDML, VoiceXML,

Description

A VoiceXML Dialogue apparatus based on Speech Act for Controlling Dialogue Flow and method of the same}

1 is a diagram illustrating an example scenario according to a weather search dialog area in a conventional VoiceXML dialog system.

2 is a diagram showing the configuration of a speech act based VoiceXML dialogue apparatus for dialogue flow control according to the present invention;

3 is a diagram illustrating a part of DDML in the configuration of a VoiceXML dialog according to the present invention.

Figure 4 is an embodiment showing the DTD of the DDML in the configuration of the VoiceXML chat device according to the present invention

5 is a flowchart illustrating a dialogue act based VoiceXML dialogue method for dialogue flow control according to the present invention.

6 is a diagram illustrating a part of DDML reflecting multiple conversation flows in a VoiceXML dialogue method according to the present invention;

7 is a diagram illustrating a part of speech XML based voiceXML in a VoiceXML dialogue method according to the present invention;

8 is an embodiment of a dialogue act-based VoiceXML dialogue method for dialogue flow control according to the present invention.

* Explanation of symbols for main parts of the drawings

100: VoiceXML chat unit 110: chat manager

112: speech recognizer 114: conversation manager

116: speech synthesizer 120: VoiceXML interpreter

200: offline part 210: web server

212: VoiceXML document 220: DDML document

230: Conversation scenario 240: DDML2VoiceXML

250: Scenario2DDML

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the field of voice conversation interfaces, and more particularly, to a speech act-based VoiceXML (Voice Extensible Markup Language) conversation apparatus and method for dialogue flow control.

The VoiceXML is a voice-based Internet standard language. If the Internet-based service is implemented by providing a web-based service scenario on a PC through an HTML language, the corresponding voiceXML is an interactive voice response device through the VoiceXML language. Voice Response: It is a service that is implemented on VoiceXML platform corresponding to IVR) and provided through telephone voice.

As such, the VoiceXML is basically a markup language that enables users to control data on the web using telephone voice. Currently, the VoiceXML is used in a conversation system. Has the advantage of enabling the control of the conversation flow that could not be controlled in the existing conversation system.

However, since the VoiceXML describes a conversation scenario based on the speaker's voice, there is a limitation in describing the conversation flow.

For example, if VoiceXML is described to proceed with the next conversation about the speaker's "Hello," VoiceXML will not proceed any further if the speaker's intention is the same as "Hello" but the voice is wrong.

To help understand, one embodiment according to the same will be described with reference to the drawings.

1 is a diagram illustrating an example scenario according to a weather search dialogue area in a conventional VoiceXML dialogue system.

Referring to FIG. 1, a user first calls a robot called "Sunny", and the robot responds to the user's call with "What can I help you with?" Then, the user requests a weather search again, "How is the weather in Daejeon today?", And the robot searches for the weather and provides the user with "The weather in Daejeon is clear".

However, based on this scenario, the VoiceXML system provides weather information only when the user inquires "How is the weather in Daejeon today?", And other queries such as "Please tell me the weather in Daejeon" or "Today." Do you like weather in Daejeon? "

This is because the VoiceXML document describes the dialogue itself, "How's the weather in Daejeon today?" In other words, the VoiceXML conversation system will not proceed any further conversation if it does not match the conversation contents stored in the VoiceXML document and the input from the user, "How about weather today?"

As such, the existing VocieXML system has been applied to many commercial systems because the developer can easily apply the dialogue scenario to the dialogue system even if the developer does not know the internal structure of the dialogue system. However, since VoiceXML directly describes the contents of the conversation, it has a disadvantage in that a lot of restrictions are placed on the processing of the conversation.

Therefore, the use is currently limited to the area that can be handled by the system-driven dialogue, such as various guide systems and reservation systems.

As described above, the VoiceXML (Voice Extensible Markup Language) dialog system according to the related art described above has the following problems.

First, the conventional VoiceXML dialogue system has a problem in that it does not provide flexibility in proceeding with the dialogue because it defines the dialogue flow based on the speaker's voice.

Second, the conventional VoiceXML chat system has a problem that the chat flow is already defined and the chat flow in the chat area is not easily changed.

Accordingly, an object of the present invention is to provide a VoiceXML dialogue apparatus and method capable of freely controlling a conversation flow by applying VoiceXML and Dialogue Description Markup Language (DDML).

Another object of the present invention is to provide a VoiceXML dialogue apparatus and method for easily modifying and changing a dialogue flow through VoiceXML converted from DDML described based on a dialogue act.

Still another object of the present invention is to enable a developer to manage conversation flow more flexibly by configuring conversation management and conversation flow management independently.

Features of the dialogue act-based VoiceXML dialog apparatus for controlling the conversation flow according to the present invention for achieving the above object is a dialog manager for processing the dialog management according to the dialogue content for the utterance of the input speaker, and DDML (Dialogue Description Markup) And a VoiceXML module for controlling the dialogue flow of the dialogue manager based on the dialogue act information extracted from the dialogue contents through a language), wherein the dialogue manager recognizes the speaker's voice and delivers the recognition result to the dialogue manager. Analyzing speech recognition results, extracting dialogue acts, and generating dialogue sentences based on VoiceXML documents loaded from the web server through the VoiceXML module together with extracted dialogue acts, and synthesizing the synthetic sentences generated by the dialogue manager. It further comprises a speech synthesizer for responding to the speaker.

delete

Preferably, the VoiceXML module uses Scenario2DDML to generate a dialogue scenario extracted from a dialogue DB of a specific region as a DDML document in an offline manner corresponding to the dialogue act information extracted from the dialogue contents through the dialogue manager, and based on the dialogue DDML document. Scenario2DDML which converts VoiceXML document into a web server and stores it in a web server, and a speech manager markup language of a speech act type VoiceXML that is loaded and processed based on a dialogue act transmitted from the dialogue manager. It comprises a VoiceXML interpreter to deliver.

Preferably, the generated DDML document may further include a DDML editor which complements the conversation flow required by the developer or controls the detailed conversation flow.

Preferably, the DDML is configured to configure each conversation in a state unit including a speech object (<object>), a dialogue act (<action>), and a target (<target>).

To achieve the above object, a feature of the dialogue act-based VoiceXML dialogue method for dialogue flow control according to the present invention includes (a) recognizing a speaker's voice and outputting a recognition result when the speaker conceives; (B) extracting dialogue act information from the dialogue contents by analyzing the speech recognition result; (c) extracting dialogue scenarios from the dialogue DB of a specific region in an offline manner corresponding to the dialogue act information extracted from the dialogue contents; and (d) Extracting a dialogue act from the extracted dialogue scenario to generate a DDML document expressed in DDML reflecting multiple dialogue flows, and (e) converting the generated DDML document into a dialogue act based VoiceXML document and storing it in a web server. And (f) processing a conversation by loading a VoiceXML document stored in the web server and generating a response dialogue line based on the extracted dialogue act information. And, (g) characterized by comprising the step of generating a response sentence based on the generated response information and synthesized speech act in response to the speaker.

delete

Preferably, step (d) further comprises supplementing the DDML if it is determined that the expressed DDML deviates from the conversation flow required by the developer or needs more detailed control of the conversation flow.

Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments with reference to the accompanying drawings.

Referring to the accompanying drawings, a preferred embodiment of the dialogue act-based VoiceXML dialogue apparatus and method for dialogue flow control according to the present invention will be described.

2 is a diagram illustrating a configuration of a speech act-based VoiceXML conversation apparatus for conversation flow control according to the present invention.

As shown in FIG. 2, the VoiceXML conversation apparatus is largely defined as a VoiceXML conversation unit 100 and an offline unit 200. At this time, the offline unit 200 is a block for providing information necessary for the control of the conversation flow on the offline, and the actual VoiceXML dialog device performs all operations in the VoiceXML dialog unit 100.

The VoiceXML conversation unit 100 is a conversation manager 110 that processes conversation management according to a conversation content such as recognition of a speaker's utterance, intention grasping and output through synthesis, and the conversation manager 110. It consists of a VoiceXML interpreter 120 for delivering a response response corresponding to the speech act.

In addition, the conversation manager 110 analyzes a speech recognizer 112 that recognizes a speaker's voice, extracts a speech act by analyzing a speech recognition result, and loads it from the web server through the VoiceXML interpreter 120 together with the extracted speech act. The conversation manager 114 generates a compound sentence based on the VoiceXML document, and the voice synthesizer 116 synthesizes the compound sentence generated by the conversation manager 114 and responds to the speaker. Accordingly, the VoiceXML interpreter 120 loads and processes the markup language of SpeechXML VoiceXML, which is processed by loading the VoiceXML document 212 stored in the web server 210 based on the dialogue act transmitted from the conversation manager 114. Transfer to conversation manager 114.

In addition, the offline unit 200 generates a dialogue scenario extracted from the dialogue DB of a specific region as a DDML document 220 in an offline manner corresponding to the dialogue act information extracted from the dialogue contents through the VoiceXML dialogue unit 100. ), And the Scenario2DDML 250 that converts the generated DDML document 220 into a speech act based VoiceXML document 212 and stores it in the web server 210.

In this case, if the generated DDML document 220 deviates from the conversation flow required by the developer or needs more detailed control of the conversation flow, the DDML document 220 may be supplemented with a DDML editor (not shown).

3 illustrates a weather search scenario as an example of a part of the DDML in the configuration of the VoiceXML dialog according to the present invention.

As shown in FIG. 3, the DDML includes each conversation in state units, and each state again includes a utterance object (<object>), a dialogue act (<action>), and a target (<target>). Other additional information is included.

As described above, the DDML is a markup language for describing a conversation flow based on a dialogue act, and the developer automatically uses the DDML document from the dialogue scenario 230 according to the document type definition (DTD) of the DDML indicated by 4. 220). Therefore, the developer can easily change and modify the conversation scenario by using the DDML in an offline manner.

The operation of the dialogue act-based VoiceXML dialogue method for dialogue flow control according to the present invention configured as described above will be described in detail with reference to the accompanying drawings.

5 is a flowchart illustrating a speech act based VoiceXML dialogue method for dialogue flow control according to the present invention.

Referring to FIG. 5, first, when a speaker thinks, the speech recognizer 112 recognizes the speaker's voice and transfers the recognition result to the conversation manager 114.

Subsequently, the conversation manager 114 analyzes the voice recognition result to extract speech act information and transfers the extracted speech act information to the VoiceXML interpreter 120.

Then, the VoiceXML interpreter 120 loading the VoiceXML document stored in the web server proceeds with the dialogue based on the dialogue act and the VoiceXML document received from the dialogue manager 114, and displays dialogue act information corresponding to the response sentence. (114).

At this time, the VoiceXML document 212 is generated through the following process based on speech acts.

First, the conversation scenario is extracted from the conversation DB of a specific area.

Then, the dialogue act is extracted from the extracted dialogue scenario through Scenario2DDML 126 to generate a DDML document 220 expressed in DDML reflecting multiple dialogue flows based on the DDML Document Type Definition (DTD) as shown in FIG. 4. At this time, if the expressed DDML is out of the dialogue flow required by the developer or if more detailed control of the dialogue flow is required, the developer can supplement the DDML document.

The modification of the DDML document is necessary here to make the conversation more flexible. That is, since the conversation scenario is generally generated by extracting from the conversation DB, it is highly likely that only the general conversation flow will be described. Therefore, by loading DDML, exception handling and area-dependent dialog processing are required.

An embodiment showing a part of the DDML reflecting the multiple conversation flow is shown in FIG. 6.

The DDML document 220 generated as described above is converted into a speech act based VoiceXML document 212 through the DDML2VoiceXML 124 and stored in the web server 210.

An embodiment showing a part of the speech act based VoiceXML is illustrated in FIG. 7.

The DDML is a markup language for describing a conversation flow based on a dialogue act. The DDML generates a VoiceXML through a DDML from a conversation scenario in an offline manner. The developer can easily change and modify a conversation scenario using the DDML. .

In the following description, the conversation manager 114 generates a response sentence based on the dialogue act information transmitted from the VoiceXML interpreter 120 and transmits the generated response sentence to the voice synthesizer 116.

Finally, the speech synthesizer 116 synthesizes a response sentence and responds to the speaker.

FIG. 8 is a diagram illustrating multiple dialogue flows based on a weather search as a dialogue act as an embodiment of a dialogue act based VoiceXML dialogue method for dialogue flow control according to the present invention.

Referring to FIG. 8, when the speaker calls the robot, the conversation manager 110 extracts the dialogue act information (system_call) extracted from the conversation contents and transmits the dialogue act information (system_call) to the VoiceXML interpreter 120 (S100).

The VoiceXML interpreter 120 returns a call_response to the conversation manager 110 and then waits for information on a conversation of the next conversation (S200).

And when the speaker utters "tell me the weather of Daejeon today", the conversation manager again extracts a dialogue act (search_weather_date_place) corresponding to the weather search (S300).

In this case, the DDML should be able to describe multiple conversation flows because in the actual conversation method, the conversation flow is branched into several conversation flows according to the user's response.

For example, in the weather search dialog, as in Figure 1, the user asks, "How is the weather in Daejeon today?" You can query the date and region according to the weather (search_weather (date, place)), but you can say "How is the weather today?", "How is the weather in War?", Or "Will you tell me the weather?" As such, a query based on weather including only a date (search_weather (date)), a weather based query including only a region (search_weather (place)) or a weather based query without a date and region (search_weather) may be performed.

As such, each utterance has the same intention as the speaker, but because the information contained in the utterance is different from each other, they must have different conversation flows.

Accordingly, in order to describe such a multi-conversation flow, DDML defines elements and attributes for dialog branch control such as if, switch, goto, link, etc. as shown in FIG.

Therefore, the VoiceXML interpreter 120 loads the VoiceXML document converted from the DDML document in which the multi-conversation flow is described, processes the corresponding conversation, and returns the response line to the conversation manager 110 (S400).

The conversation manager 114 which has returned the response dialogue line generates a response sentence corresponding to the response dialogue line and transmits the generated response sentence to the voice synthesizer 116.

Through the dialogue act-based VoiceXML, it is possible to control the dialogue flow more flexible than before. In other words, in case of existing VoiceXML, if the voiceXML describes "tell me the weather today", the speaker can speak only when "tell me the weather today." Is it nice today? " It is possible to control the conversation flow more flexibly, because the act of speech can have the same variety of voices.

As described above, the preferred embodiment of the present invention has been disclosed through the detailed description and the drawings. The terms are used only for the purpose of describing the present invention and are not used to limit the scope of the present invention as defined in the meaning or claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the dialogue act-based VoiceXML dialogue apparatus and method for dialogue flow control according to the present invention have the following effects.

First, the speech act based VoiceXML device and method according to the present invention can handle the conversation area that the existing VoiceXML cannot process, so that the user can feel more intimate in the conversation with the system, and the application field can also use the existing area. More broadly applicable.

Second, in the present invention, by managing only the management of the conversation flow by VoiceXML, it is possible to configure the conversation management and the management of the conversation flow (conversation scenario) independently so that the developer can manage the conversation flow more flexibly.

Claims (8)

A conversation manager that handles the conversation management according to the conversation contents of the speaker's voice input; It includes a VoiceXML module for controlling the dialogue flow of the dialog manager based on the dialogue act information extracted from the dialogue contents through DDML (Dialogue Description Markup Language), The conversation manager recognizes the speaker's voice and delivers the recognition result to the conversation manager. A conversation manager for generating a composite sentence based on the speech synthesizer, and a speech synthesizer responding to the speaker by synthesizing the synthesized sentence generated by the conversation manager Speech act based VoiceXML dialogue device further comprising. delete The method of claim 1, wherein the VoiceXML module Scenario2DDML for generating a DDML document of a dialogue scenario extracted from a dialogue DB of a specific area in an offline manner corresponding to the dialogue act information extracted from the dialogue contents through the dialogue manager; Scenario2DDML converting the generated DDML document into a speech act based VoiceXML document and storing it in a web server; Speech act based VoiceXML dialogue device including VoiceXML interpreter which delivers markup language of speech act type VoiceXML to conversation manager based on dialogue act delivered from dialogue manager. . The method of claim 3, wherein A dialogue act-based VoiceXML dialogue device further comprising a DDML editor in which the generated DDML document deviates from the dialogue flow required by the developer or through the control of the finer dialogue flow. The method of claim 3, wherein The DDML is a dialogue act-based VoiceXML dialogue apparatus that configures each conversation in a state unit including a speech object (<object>), a dialogue act (<action>), and a target (<target>). (a) when the speaker thinks, recognizes the speaker's voice and outputs the recognition result; (b) analyzing speech recognition results to extract speech act information from conversation contents; (c) extracting a dialogue scenario from a dialogue DB of a specific region in an offline manner corresponding to the dialogue act information extracted from the dialogue contents; (d) extracting a dialogue act from the extracted dialogue scenario to generate a DDML document expressed in DDML reflecting multiple dialogue flows; (e) converting the generated DDML document into a speech act based VoiceXML document and storing it in a web server; (f) processing a conversation by loading a VoiceXML document stored in the web server and generating a response dialogue line based on the extracted dialogue act information; and (g) generating and synthesizing a response sentence based on the generated response dialogue information and responding to the speaker. delete The method of claim 6, The step (d) further comprises the step of supplementing the DDML if it is determined that the expressed DDML deviates from the conversation flow required by the developer or needs more detailed control of the conversation flow.

Images