KR100488112B1 - Apparatus For Converting Document and Searching in Voice Portal System - Google Patents

Abstract

The present invention relates to an apparatus for presenting a search result in a voice portal system.

Therefore, the present invention provides an indexer for indexing index in a VXML-based speech portal system, an information search engine that matches a keyword or query word with an index word, and finds a URI (Uniform Resource Indentifier) of a keyword's destination or scenario, and a menu based on the search results. Adds a scenario generator that automatically creates and executes scenarios that can provide additional services such as moving to or automatic call connection, and provides a kind of document processing function to quickly and easily select VXML documents and systematically store and manage scenarios. do.

Therefore, when the user makes a phone call and speaks a keyword of information, the user can listen to the corresponding information on the Internet by voice, and can provide a voice portal service capable of automatic call connection.

Description

Apparatus For Converting Document and Searching in Voice Portal System

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search and result presentation apparatus in a voice portal system, and more particularly, to an apparatus for searching for a service and presenting the results so that a user can use the voice portal system more quickly and conveniently.

In extending the conventional text-based internet environment to the voice internet environment, a system built using the nickname markup language called Voice eXensible Markup Language (hereinafter referred to as VXML) provides an advanced level of voice service. Expectations are the most powerful way to realize this.

VXML is a type of XML document designed for voice application development. Here, the voice application means that a voice-based user interface such as digitized audio, voice recognition, DTMF key input, voice input, telephone, or the like can be used in computer telephony integration (CTI), WEB, and DB. The VXML technology was developed by three companies, AT & T, Lucent Technologies, and Motorola, and is a voice-based technology that can search various information on the Internet through telephone and voice recognition software. In addition, VXML is a technology for retrieving contents such as e-mail, weather information, traffic information, etc. of the Internet through a telephone without an Internet access device such as a PC or a notebook, and loading various voice information on a web page. At the present time when technologies such as speech synthesis, recognition, and authentication are put into practical use, voice information services using VXML technology are spreading.

With the use of such VXML, existing voice service providers can easily provide a variety of new services, and the division of work on voice service authoring clearly occurs.

In the past, in order to provide a voice service, a service author had to work on a technical scenario regarding voice input and output along with a real scenario. However, with the use of VXML, service authors only need to author their services according to the document format, and the remaining voice related technologies are automatically handled by the VXML browser.

VXML is based on XML, but the characteristics of existing web documents and VXML documents are similar, using tags only.

Therefore, construction of new content or conversion of existing content into a new form that makes good use of the characteristics of voice media is inevitable, and a conventional VXML-based voice portal system for this purpose is shown in FIG. 2 as a VXML interpreter (VoiceXML Interpreter; 230, a VXML Interpreter Context 210, an Implementation Platform, and a Document Server 220.

Referring to FIG. 2, the VXML interpreter 230 is the most essential element of an execution environment, and is responsible for loading a VXML document, interpreting the contents, and executing the content. That is, according to the functions set in 47 kinds of tags, the sequential flow of document execution is controlled, the voice input / output contents are determined, and the commands required for the voice platform are issued.

The VXML Interpreter Context 210 is responsible for controlling and managing the VXML Interpreter 230 according to the system-wide environment. In particular, in case of multi-channel interpreter for ASR service or voice portal support, it detects user access and release and controls the allocation, start, and stop of interpreter thread, and manages start document specification, environment variable and system option setting. do. In other words, it is an outer element set to maximize the interpreter's environmental independence.

The document server 220 transmits a document or resource requested by the VXML interpreter 230 transmitted in the form of a Uniform Resource Identifier (URI) to the interpreter. In case of document transfer request by submit structure, voice file or grammar file is transmitted in case of audio or grammar structure.

The implementation platform supports voice input / output and call management while being controlled by the VXML interpreter 230 or the VXMLL interpreter context 210. The audio output function supports audio file playback, TTS speech synthesizer, DTMF button input, voice recognition by ARS function, and audio input recording.

Referring to Figures 1 and 2 the operation of the prior art device will be described.

First, telephone users 1 and 2 on telephone network 10 dial a specific number to receive voice portal service.

Then, the session manager 212 of the VXML server 20 receives a phone call, establishes a new session, and starts an event handler 215 to handle subsequent events.

The document management of the VXML server 20 obtains a scenario (VXML document) desired by the user from the document server (DOC Server) 220 and delivers it to the interpreter 230.

The interpreter 230 of the VXML server 20 parses the VXML document and finishes preparing to execute the scenario.

When the speech synthesis for the sentence is required in the scenario, the interpreter 230 requests the event handler 215.

The event handler 215 performs the speech synthesis through the TTS interface 216 and passes the result to the telephone interface 211 so that the user can listen through the telephones 1 and 2. To help.

When a request for speech recognition occurs, the interpreter 230 requests the event handler 215.

The event handler 215 prepares to recognize this through the ASR interface 216 and receives a result of recognizing a voice received through the telephony interface 211 from the ASR 260. This is returned to the interpreter 230.

The interpreter 230 proceeds to the next service according to the recognition result and repeats the above process.

When the user hangs up, the telephone interface 211 notifies the event handler 215 and finally the service is terminated by the session manager 212 closing the session.

Users who have become accustomed to the Internet these days have increased their appetite for information, but have lost their patience. In other words, in the hierarchical menu structure that provides most voice services, users are burdened with the process of listening to and passing the guidance message by the waiting time or step required to move to the lower menu. This is due to the charges that are applied in proportion to the usage time, but also due to the propensity of users who dislike complicated and prefer direct and immediate response.

Accordingly, an object of the present invention is to provide a search and result presentation device so that general users can use the voice portal system in a VXML-based environment more quickly and easily, and at the same time, provide a base for service providers to create new services more freely. .

In order to achieve the above object, the present invention provides an information retrieval engine that finds a URI (Uniform Resource Indent) of a keyword's destination or scenario by matching an indexer for indexing with an existing VXML-based voice portal system and a keyword or a query and an index. Retrieval Engine) and a Scenario Generator that automatically creates and executes scenarios that can provide additional services such as navigation to menus or automatic call connections based on the search results. It provides a kind of document processing function to save and manage scenarios.

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

3 is a structural diagram of a VXML-based voice portal system capable of voice search and automatic call connection according to the present invention, and is configured by adding an information retrieval server 300 in the existing voice portal system.

The present invention is an indexer (320) for indexing the index, such as shown in Figure 3 in the existing VXML-based speech portal system, match the keyword or the query and the index word, the URI (Uniform Resource Indentifier) of the destination or scenario of the keyword A search engine (Retrieval Engine) 310 for finding a search engine and a scenario generator (330) that automatically creates and executes a scenario that can provide additional services such as moving to a menu or an automatic call connection based on the search results. In addition, it provides a kind of document processing function for selecting VXML documents quickly and easily and systematically storing and managing scenarios.

The search system in the VXML-based speech portal system according to the present invention supports two kinds of searches: keyword search and content-based search. Keyword search is a service that the user knows the contents of the service and searches directly without hierarchical menu paths for each node and extends to automatic call connection. And, content-based search is a service that finds and executes a suitable scenario based on the similarity between the content provided by the VXML document and the query.

The indexer 320, the information retrieval engine 310, and the scenario generator 330 according to the present invention have different operations according to a search purpose of a keyword search or a content-based search.

The indexer 320 uses the XML Parser 232 of the VXML Interpreter 230 to index the document structure, the word to be searched in the document, and the destination.

First, keyword search is as follows.

There is one index file for each scenario of the keyword hierarchy. The perfect hierarchy does not allow any logical loop structure except the loop structure due to service error or user input error. Refer to the scenario of rescue. This means that the indexer 320 must artificially release the loop structure through the loop detection method for each node.

Index words, which are words that are subject to speech recognition, correspond to user selectable menus in a scenario, which are defined by tags (grammar, choice, filed, menu) in a VXML document.

The destination ID means an identifier of a service to be provided for a destination mapped to a menu, that is, a menu selected by a user by voice. Every VXML document has a unique id on the Internet, and elements such as forms, blocks, and menus also have unique ids in one scenario, so the destination ID is given in the form of "VXML document_URI # Element_ID" with the index word. Store in an index file.

Meanwhile, the content-based search is as follows.

Document structure does not go through separate document structure indexing process like keyword search because one index file exists for all sites provided by CP.

Not only the object of speech recognition but also the information provided as the contents of the service are indexed, which analyzes tags (grammar, choice, menu, block, prompt) and extracts them with index words.

 The scenario that contains the index words, that is, the URI of the current VXML document being parsed, becomes the destination ID of the index word. Unlike keyword search, the index file that stores the index word and its destination ID is an Inverted file [William B. Frakes, "Information Retrival", prentice Hall 1992, pp. 377 ~ 380].

When the voice is recognized through the recognizer (ASR Engine) 260, as a result, the word is transmitted in a text form, and the recognized word becomes a query of the information search engine 310. The detailed search process is performed as follows according to the purpose of the search.

First, the keyword search finds the index word and the destination ID by searching the index word that matches perfectly with the query word in the index file.

Meanwhile, in content-based retrieval, the information retrieval engine 310 calculates the similarity between the query word and the document based on the inverted index file generated during the indexing process, and selects candidate documents suitable for the query word. List them in high order.

Similarity calculation method used in the present invention is similarity calculation method based on Inverted File used in general content-based search engine [William B. Frakes, "Information Retrival", prentice Hall 1992, pp. 370–375].

Scenario generator 330 is composed of a new VXML document to serve the user the index word and the destination determined by the search process among the index words defined in the index file.

First, keyword search is as follows.

When the destination ID of the index word is a VXML document or a specific element in the document, the newly formed VXML document forms a document as follows so that service connection is automatically made.

<goto next = "VXML document_URI # Element_ID">

If the destination ID is an external phone number, the document is configured as follows to support automatic call connection.

<transfer name = "Retrieval_Result" dest = "phone: // 028181234" connecttimeout = "30s" bridge = "true">

<filed>

<assign name = 'R_res_dur' expr = "Retrieval_Reseult $ .duration" />

<if cond = "'Retrieval_Result' = 'busy'" />

<elseif cond = "'Retrieval_Result' = 'noanswer'" /> </ if> </ filled>

</ transfer>

On the other hand, in content-based search, a plurality of documents are searched in the order of high similarity as shown in web search. In order to convey this to the user, the scenario generator 330 constructs the following VXML document to allow the user to select an appropriate site. In the following rank2 means that a document whose telephone number is a destination ID is retrieved.

<form id = "Retrieval_Result">

<block> <prompt> Listen to the search results and select to move the scenario. </ prompt>

<goto next = "# List" />

</ block>

</ form>

<menu id = "List">

<choice next = "# rank1"> Index Information for Document 1 </ choice>

<choice next = "# rank2"> Index Information for Document 2 </ choice>

<catch event = "nomatch noinput"> <exit /> </ catch>

</ menu>

<form id = "rank1">

<goto next = "uri in document1">

</ form>

<form id = "rank2">

<transfer name = "Retrieval_Result" dest = "phone: // 028181234" connecttimeout = "30s" bridge = "true">

<filed>

<assign name = 'R_res_dur' expr = "Retrieval_Reseult $ .duration" />

<if cond = "'Retrieval_Result' = 'busy'" />

<elseif cond = "'Retrieval_Result' = 'noanswer'" /> </ if> </ filled>

</ transfer>

</ form>

As described above, the present invention can provide information by searching and converting XML-type information into voice in a VXML-based voice portal system.

Therefore, when the user makes a phone call and speaks a keyword of information, the user can listen to the corresponding information on the Internet by voice, and can provide a voice portal service capable of automatic call connection.

In addition, the present invention can save the waiting time or listening time of the guidance message for the user to move to the lower menu in using the service.

In addition, when the VXML-based voice portal service is expanded and generalized in the future, the present invention enables the service to search the Internet, listen to the result by voice, and move to a scenario of a specific service provider.

1 is a diagram showing a connection relationship for providing a voice mail service.

2 is a block diagram of a conventional voice portal system.

3 is a block diagram of a voice portal system according to the present invention.

Claims (4)

In an apparatus for searching for a service in a voice eXensible Markup Language (VXML) -based voice portal system and presenting the result, An indexer for indexing indexing; An information search engine that matches a keyword or query word with an index word and finds a Uniform Resource Indent (URI) of the destination or scenario of the keyword; An information retrieval server is provided that includes a scenario generator that automatically creates and executes a scenario that can provide an additional service such as moving to a menu or an automatic call connection based on the retrieved result. The search server supports two types of searches, keyword search and content-based search, and operations of the indexer, information search engine, and scenario generator are performed differently according to the search type. And document conversion device. The method of claim 1, wherein the indexer, An apparatus characterized by indexing a document structure, a search target word and a destination in a document by using a VMS document analyzer of the VMS interpreter. The method of claim 1, wherein the information search engine, And calculating candidate similarities between the query word and the document based on the inverted index file generated during the indexing process, and listing candidate documents suitable for the query word in the order of high similarity. The system of claim 1, further comprising: a scenario generator; An apparatus comprising a new VXM document to provide a service to a user, the index word determined by a searching process among the index words defined in the index file, and a destination thereof.