KR100497543B1 - Apparatus for converting HTML to VoiceXML having a selecting function of transformation purpose by prior knowledge and communication of user and method for employing as the same - Google Patents

Abstract

The present invention relates to an HTMLtoVoiceXML converter having a function of selecting a conversion target through interaction with a user's prior knowledge and interaction for converting a content of a web document written in HTML into a VoiceXML document to provide a voice interface. Enter the URL of the web page you want to convert to automatically interact with the user on the web page or to the desired content group based on a predefined rule base, regardless of the implementation method of the original web document. It provides an automatic conversion method that selects and separates the contents for the corresponding contents to generate and save the VoiceXML document according to the voice interface scenario. In addition, a continuous web document in which a list of contents has the same form and structure information over several pages can be accessed in a single access by selecting a structural feature of a graphic content or graphic for connection and connection of the first web page. This allows conversion of all linked web pages. The present invention defines the conversion target required by the user by using text, rules, and visual expression among contents in the HTML web document, so that accurate target content can be selected regardless of various authoring techniques. It can easily detect the personality change of and make it possible to reliably decide whether to select the content according to the change of web page composition that can appear over time after user selection. In addition, by automating the work after the user selects a conversion target, the cost and time required to convert a large-format HTML web document into a VoiceXML document can be greatly reduced.

Description

Application for converting HTML to VoiceXML having a selecting function of transformation purpose by prior knowledge and communication of user and method for employing as the same}

The present invention relates to an HTMLtoVoiceXML converter and its operation method, in particular HTMLtoVoiceXML having a conversion target selection function through the user's prior knowledge and interaction for converting web information written in HTML to VoiceXML format to provide a standardized voice interface The converter and its operation method.

The most common method of converting HTML to other forms of markup languages is to replace tags of other markup languages, which are similar to the functions of HTML tags, and are often used frequently when converting HTML to markup languages for wireless Internet. It became. However, since the use of the two markup languages is different, the conversion of only the tags is awkward in the construction of the information delivery scenario, so that the information is not clearly delivered. Unlike the markup language for the wireless Internet, VoiceXML tags are written to provide information through a voice interface. Therefore, since the purpose is different from HTML tags written for visual information, it is impossible to replace the tags. The way the information delivery scenarios are organized is also different. Also, in general, users do not use all the content displayed in the HTML document, but only the necessary content. As a method of selecting the necessary contents from a web page, conventionally, the structure analysis method such as converting the web page into a tree structure and selecting a tree having the most child nodes as the content as the content, and ontology as a content-based approach The method of selecting contents with high suitability by describing the contents and relations of information required by (ontology) has been used, but in the case of the structural method, there is a possibility of selecting contents different from the intention of the user. In some cases, the content is visually the same, but is represented in a different tree, and according to the implementation technique, the user may have a structural characteristic different from the content group recognized by the visual representation of the web page. The content-based approach is a good way to select only the content that the user needs on the Internet.However, the content-based approach is complicated to select the necessary content for the web page that has been decided. Requires expertise and complexity to be analyzed. In addition, the structure and contents of the web document is changed from time to time, which is inconvenient to update the ontology document at each time. Therefore, it requires less expertise and does not match the characteristics of the converter used to reduce the complexity and the time required for writing the document. Conventional converters have been limited to web pages connecting the target, which is inconvenient to undergo a number of conversion process for information conversion of a website having a large amount of information of the same nature and structure.

An object of the present invention is to read and convert a web document written in HTML into a tree structure to form a group of related content, extract the URL of a continuous linked document having the same content from the read HTML web document, The present invention provides an HTMLtoVoiceXML converter having a function of selecting a conversion target through interaction with the user's prior knowledge and a method of presenting the content group information to the user to determine the user's selection and setting the target to be converted into a VoiceXML document. .

In order to reliably select the content group in the HTML document as needed, the characters included in the web document are used, or a visual representation of the web page corresponding to the tree information generated by the preprocessing unit of the converter is provided separately, or generated. The organization method and implementation method using HTML of the web page compared to the conventional method by using the method of sorting the target content group through interaction with the user who selects the number of nodes and characters in the tree based on rules Regardless, you can make choices that are reliable, accurately reflect your intentions, and make multiple choices on one web page.

It not only separates, extracts and stores the list of contents in the intelligent and heuristic way, but also extracts and saves the contents through the corresponding link at the same time. Handles web pages.

In order to provide information through voice, we reconstruct information delivery scenario and generate VoiceXML document based on this.

Also, depending on the content, such as newspapers and job sites, the list consists of a large number of web pages that have the same personality and structure and are linked in succession. If you do not extract all of the content and reflect it in constructing an information delivery scenario, you will not be able to communicate clearly.

When the content is composed of multiple documents, since consecutive documents are linked to each document, it is inefficient to extract the content by accessing each document individually to extract the content.

However, since HTML code alone does not know which link is a link to a continuous linked document, the user can obtain information about the link relation to the linked linked document through interaction with the user, and extract content from multiple documents with only one access. Do it.

The present invention is a converter that converts web information written in HTML into VoiceXML so that web information can be accessed through a voice interface, and converts the content to be converted by using DB prior knowledge or interaction with the user. By selecting and converting only the information required by the user, a clear information transmission scenario can be configured, and the code can be written concisely and clearly by preventing the conversion of unnecessary information. In addition, content that spans multiple web documents with the same characteristics can be extracted with only one connection, thereby increasing the efficiency of the converter.

Fig. 1 shows the structure of the HTMLtoVoiceXML converter according to the present invention. The list extracting / VoiceXML generating unit 101 connects a web page, selects a group of contents of the connected web page, and separates and extracts a list of contents. And a detail content extraction / VoiceXML generation unit 102 for extracting detailed content linked by a list of content and generating the content in VoiceXML, and for continuous web document processing having the same personality and structure. It is composed of three parts of the connection document URL extraction unit 103.

When the URL extraction 104 of the HTML document to be converted is input, the list extraction / VoiceXML generation unit 101 accesses a web page of a website corresponding to the URL, and the corresponding web page through the list HTML analysis module 105. Extracts the list content group information from the HTML document, and transmits the list content group information to the list content group selection module 106. The list content group selection module 106 shows the user a screen divided into the list content group to separate the list content group. Allows the user to select the list of conversion target list contents through the screen, or automatically selects the list content group containing the selected character string by selecting and copying and pasting text information included in the web document, Allows the user to select a list content family based on a set of rules, such as a tree with the largest number of nodes It provides a method and to thus save the selected content list information group of webpage listed in the content group information DB (110).

When the user selects the list content group, the content may be represented as one content group by merging two or more content groups. After the content group selection process is completed, the content is extracted through the list content extraction module 107 for the selected list content group. If there is link information in each of the extracted list contents, the link information (link URL, link text) is extracted from each content link information extracting module 109 and linked to the detail content extracting / VoiceXML generator 102 to extract the detail contents. Passes URL and link text information. The VoiceXML document for the list content group is generated through the list VoiceXML document generation module 108 for all contents extracted through the list content extraction module 107. In this case, the number of generated List VoiceXML documents is generated according to the number of extracted list contents and the average size of strings of the list contents. The generated VoiceXML document is stored in the web server 119.

The link information delivered through each content link information extraction module 109 extracts the detailed content group information through the detailed content HTML analysis module 111 to extract the detailed content, and then returns to the detailed content group selection module 112. The detailed content group information selected through the module is delivered to the detailed content content extraction module 113 to extract the detailed content content, and the detailed content voice XML generation module 114 targets the extracted content. Create a VoiceXML document for the content. Details The VoiceXML document is generated as one document for the details, and the generated VoiceXML document is stored in the Web Server 119.

If the list contents exist in several documents having the same personality and want to extract contents in other documents other than the current document, the prior knowledge input module 118 receives the connection relationship information of the continuous connection document from the user through the prior knowledge input module 118. Save to the DB 117, save the HTML document at the first input URL through the HTML document storage module 115, deliver the code content of the HTML document stored to the continuous link document URL extraction module 116, and The connection document URL extraction module 116 analyzes prior knowledge and HTML codes stored in the prior knowledge DB 117, extracts a continuous connection document URL, and delivers the URL to the list extraction / VoiceXML generator 101 to extract the list contents. .

FIG. 2 is a flowchart of the list HTML analysis module 105. When a URL is input, the website is accessed 201, the HTML document is extracted 202, and a document object is converted into a tree structure 203. The content group information is extracted by sequentially analyzing the root node (204). First, a tag name of each node is obtained to determine whether or not it is an analysis target (205). Whether to analyze the tag improves the analysis efficiency by excluding the tag name as comment (!) Tag, SCRIPT tag, and STYLE tag from the analysis target. If it is an analysis target, it checks whether the number of child nodes of the current node is three or more (206), and if it is three or more, extracts a subtree rooted at the current node (207) and checks whether an internal string exists (208). do. If the internal string exists, the name and index of the current node and the last node of the subtree are obtained (209) and stored in the content group information array (210). Then, the pointer of the current node is moved to the next node (211), and it is repeated from the process of determining whether the node is the analysis target until the node pointer is the end of the tree (212). If it is the end of the tree, the operation ends (213).

3 is a flowchart of the list content group selection module 106. The content group information 300 is received through the list HTML analysis module 105, and the number is displayed 301 for each content group of the web document to the user. The user selects the content group to be converted by the number displayed on the screen. If two or more content groups are similar in this process, the number of each content group is selected and merged as a merge target (302). Then, a title appropriate to the selected content group is set (303). The selection operation is repeated until the content group selection operation is finished (304).

4 is a flowchart of the details HTML analysis module 111. When the link URL and the link text information 400 extracted by each content link information extraction module 109 of FIG. 1 are delivered, the corresponding website of the link URL is transmitted. In operation 401, the HTML document is extracted 402, converted into a tree structure through a document object 403, and sequentially analyzed from the root node 404 to extract content group information. The analysis process first determines the tag name of each node to determine whether it is an analysis target (405). Whether to analyze the tag improves the analysis efficiency by excluding the tag name as comment (!) Tag, SCRIPT tag, and STYLE tag from the analysis target. If the target is to be analyzed, the number of child nodes of the current node and the tag names of the child nodes are obtained (406), and if more than half of the number of child nodes is a BR or P tag (407), and if correct, the subtree rooted at the current node is extracted ( In operation 408, the current node name, index, and text size of the subtree are obtained (409) and stored in the content group arrangement (410). Then, the pointer of the current node is moved to the next node (411), and the process is repeated from the analysis target determination process of the node until the node pointer is the end of the tree (412). If it is the end of the tree, the operation ends (413).

Depending on the content, there is a marked as a continuous linked document of the same type. In order to extract all of these types of content at once, it is necessary to obtain the URL of the continuous linked document and extract the content from the document. However, since the HTML code alone does not know which link is a link to the continuous link document, the link information is based on the input information received from the user, such as the link type of the continuous link document and the relative position of the last content. Can be extracted.

FIG. 5 is a diagram representing a schema for receiving prior knowledge of link information of a continuous connection document from a user, and is largely composed of three steps. The information input method, like the wizard in Windows, allows the user to select options step by step so that information can be easily obtained from the user. Step 1 of prior knowledge input information designates (502) what is the form of a linked link of a continuous linked document displayed in the document. There are three types of link links in a continuous link document. The first is text such as “next” or “previous”, which sometimes indicates the form of a link, and the second is displayed as an image such as an arrow or triangle light. And third, 1, 2, 3,. . . It may be indicated by consecutive serial numbers as shown below. After the user selects the link form of the continuous link document of the first stage, the second stage receives the detailed information of the link link displayed in the first start document and the intermediate document. If the link form selected in step 1 is the text 503, the link text should be input 506. Input of the link text inputs the link text 507 of the first starting document and the link text of the intermediate document 508. ) When the connection link type is an image 504, a relative link position is received 509 based on the last content. In this case, the link position 510 displayed in the start document and the link position 511 displayed in the intermediate document are separately input. If the connection link type is a serial number (505) is received (512) whether the order of the number in ascending or descending order and the start number. Finally, in step 3, up to the number of documents to be extracted, that is, the position of the last document is designated (513).

As described above, the HTMLtoVoiceXML converter having a conversion target selection function through interaction with the user's prior knowledge and interaction thereof, and a method of operating the same, convert the visual information written in HTML into a VoiceXML document to provide voice information. By selecting the content to be converted through interaction with, the user can provide only the information needed by the user, and extracts the content created by several documents with only one connection, and accesses each web document separately to extract the content. This can solve the inefficiency.

1 is an HTMLtoVoiceXML converter structure according to the present invention,

2 is a list HTML analysis module flow chart of the present invention;

3 is a flowchart illustrating a list content group selection module of the present invention;

4 is a flowchart of a detailed HTML analysis module of the present invention;

5 is a prior knowledge input schema for extracting continuous connection document URL of the present invention.

Explanation of symbols on the main parts of the drawings

101: list extraction / VoiceXML generation unit

102: extract details / VoiceXML generation unit

103: continuous connection document URL extractor

104: URL

105: List HTML Analysis Module

106: list content group selection module

107: list content extraction module

108: List VoiceXML document generation module

109: each content link information extraction module

110: list content group information DB

111: Details HTML Analysis Module

112: Contents content group selection module

113: Details content extraction module

114: Details VoiceXML Generation Module

115: HTML document storage module

116: continuous connection document URL extraction module

117: prior knowledge DB

118: prior knowledge input module

Claims (14)

As the URL of the web page to be converted by the user is entered, the user accesses the web page corresponding to the URL, extracts the list content group information from the HTML document of the web page, and stores it according to the user's selection. A list extraction / VoiceXML generation unit for extracting link information from each content and generating a VoiceXML document for list content corresponding to the selected and stored content information; By using the link information delivered from the list extraction / VoiceXML generation unit, the linked web page is accessed to extract and select detailed content group information from the HTML document of the corresponding web page, and then the detailed content is targeted to the extracted content. A detail extraction / VoiceXML generator for generating a VoiceXML document for the; And Analyze the prior knowledge entered by the user and stored in the DB, and the code of the HTML document of the web page corresponding to the URL to extract the URL of the continuously linked document, and retrieve the URL information of the continuous web document having the same characteristics and structure. Continuous connection document URL extraction unit which is transmitted to the list extraction / VoiceXML generation unit to generate a VoiceXML document for the continuous web document HTMLtoVoiceXML converter having a conversion target selection function through the user's prior knowledge and interaction, characterized in that made.  A list HTML analysis module for accessing a web page corresponding to the URL and extracting list content group information from an HTML document of the web page as a URL of a web page to be converted by the user is input; Shows the screen divided into the extracted list contents group to the user so that the user can select the conversion target list contents group through the screen divided into the list contents group, or the user copies and pastes the text information included in the web document. It provides a way to automatically select the content group that contains the selected string by selecting it through the insert process, or to select the content group according to a predetermined rule such as the tree with the largest number of nodes. A list content group selection module for storing group information in the list content group information DB; A list content extraction module for extracting content from the selected content group; A list VoiceXML document generation module for generating a VoiceXML document for a list of all extracted contents; And each content link information extraction module for transferring the link URL and the link text information to the following detail content extraction / VoiceXML generation unit to extract link information of each of the extracted list contents and extract detailed contents. VoiceXML generation unit; A detailed content HTML analysis module for extracting detailed content group information from the link information delivered from each of the content link information extraction modules; A detailed content group selection module for selecting the extracted detailed content group information; A detailed content contents extraction module for extracting detailed content contents from the selected detailed content group information; And a detail contents extracting / VoiceXML generating unit, comprising a detail VoiceXML generating module for generating a VoiceXML document with respect to the extracted contents. And A prior knowledge input module that assists a user in entering prior knowledge; A prior knowledge DB for receiving and storing connection relationship information of the continuous connection document as prior knowledge input by the user through the prior knowledge input module; An HTML document storage module for storing the HTML document at the first input URL; And extracting a continuous connection document URL by analyzing the prior knowledge stored in the prior knowledge DB and the HTML code transferred from the HTML document storage module, and extracting the list content by transferring the same to the list extraction / VoiceXML generator. Modular continuous document URL extractor HTMLtoVoiceXML converter having a conversion target selection function through the user's prior knowledge and interaction, characterized in that made. The method of claim 2, wherein when the user selects the list content group, the user selects the list content group and displays the content as one content group by merging at least two or more content groups. HTMLtoVoiceXML converter. The method according to claim 2, wherein a plurality of VoiceXML documents are generated according to the number of list contents extracted for the amount of the generated List VoiceXML document and the average size of strings of the list contents, and the generated VoiceXML documents are stored. Contents It is made by creating a single document about the details of the VoiceXML document and providing a Web Server for storing the generated VoiceXML document. HTMLtoVoiceXML converter. In the operating method of the HTMLtoVoiceXML converter having a conversion target selection function through interaction and prior knowledge of the user having the configuration of claim 2, The process of the list HTML analysis module, A first step of accessing a web page corresponding to the URL and extracting an HTML document of the web page according to the input of the URL of the web page to be converted by the user; A second step of converting the extracted HTML document into a tree structure through a document object; A third step of extracting content group information by sequentially analyzing the root node of the document; Determining a tag name of each node to determine whether to be analyzed, and if not, to proceed to the next node, and if it is to be analyzed, determining whether the number of child nodes of the current node is three or more; A fifth step of proceeding to the next node when the number of child nodes is less than three in the fourth step, and extracting a subtree rooted at the current node when the number of child nodes is three or more; In the extracted subtree, the existence of the internal string is determined. If the internal string does not exist, the process proceeds to the next node. If the internal string exists, the name and index of the current node and the last node of the subtree are obtained. A sixth step of storing; A seventh step of moving the pointer of the current node to the next node; And An eighth step of determining whether the current node pointer is the end of the tree, ending if the end of the tree, and repeating to the fourth step if not the end of the tree Operation method of the HTMLtoVoiceXML converter having a function of selecting a conversion target through the user's prior knowledge and interaction. The method of claim 5, The process of the list content group selection module, Receiving content group information from the list HTML analysis module and displaying a number for each content group of the web document to the user; Selecting a content group to be converted using a number displayed on a screen by a user; Setting a title in the selected content group; And Determining if the content group selection operation has ended and repeating the selection operation until it is finished Operation method of the HTMLtoVoiceXML converter having a function of selecting a conversion target through the user's prior knowledge and interaction. The method of claim 6, In the step of selecting the content group, if the two or more content groups are similar content group, the conversion target selection function through the prior knowledge and interaction of the user, characterized in that by merging by selecting the number of each content group as a merging target How to use the HTMLtoVoiceXML converter. The method of claim 5, The processing of the above-described details HTML analysis module, If the link URL and link text information extracted by each content link information extraction module are delivered, accessing a corresponding web site of the link URL and extracting an HTML document; Converting the extracted HTML document into a tree structure through a document object; Extracting content group information by sequentially analyzing the root node of the document; Determining a tag name of each node to determine whether it is an analysis target, and if not an analysis target, proceeds to the next node, and if it is an analysis target, obtaining the number of child nodes of the current node and a tag name of the child node; Determining whether at least half of the number of child nodes is a BR or P tag, and if less than half, proceeds to the next node, and if more than half, extracts a subtree rooted at the current node; Obtaining a current node name, an index, and a text size of a subtree, and storing the same in a content group arrangement; Moving the pointer of the current node to the next node; And Determining whether the current node pointer is at the end of the tree, ending at the end of the tree, and repeating the determination of whether to analyze the target object if not at the end of the tree Operation method of the HTMLtoVoiceXML converter having a function of selecting a conversion target through the user's prior knowledge and interaction. The method according to claim 5 or 8, If the tag name is an annotation tag, SCRIPT tag, or STYLE tag in the analysis target, HTMLtoVoiceXML converter having a conversion target selection function through prior knowledge and interaction of the user, which is excluded from the analysis target to increase the analysis speed. How to operate. The method of claim 5, In the case of the content displayed as the continuous linked document of the same format, in order to extract the content all at once, the user obtains the URL of the continuous linked document and extracts the content from the corresponding document. Operation method of HTMLtoVoiceXML converter with target selection function. The method of claim 10, The form of the link to the continuous linked document receives information of a relative position of the last content from the user, and extracts the link information based on the received information, to distinguish whether any link is a link to the continuous linked document Operation method of HTMLtoVoiceXML converter having the function of selecting conversion target through user's prior knowledge and interaction. The method according to claim 5 or 11, The process of expressing a schema that has received prior knowledge about link information about a continuous connection document from the user, Designating what type of linking links the continuous linking documents displayed on the document have; Receiving detailed information of the connection link displayed in the first start document and the intermediate document; And Steps to specify the location of the last document for which number of documents to extract Operation method of the HTMLtoVoiceXML converter having a function of selecting a conversion target through the user's prior knowledge and interaction. The method of claim 12, Specifying what the connection link type is, User's prior knowledge and interaction with the user, characterized in that the user selects any one of a series of serial numbers such as 1, 2, 3, text such as text, arrows, and triangle lights. How to operate the HTMLtoVoiceXML converter with optional features. The method of claim 12, The receiving of the detailed information may include receiving the link text of the first start document and the link text of the intermediate document when the selected link form is text, and the relative link based on the last content when the link form is an image. The location is input, but the link location shown in the start document and the link location shown in the intermediate document are input separately. If the link form is a serial number, the order of the number is ascending or descending, and the start number is received. Operating method of the HTMLtoVoiceXML converter having a function of selecting a conversion target through the user's prior knowledge and interaction.