JPH08329108A - Method for converting text into hypertext - Google Patents

Abstract

PURPOSE: To prepare a link when a part of described contents of a node is related to another node by preparing a link based upon the coincidence of stems as the meaning contents of a phrase. CONSTITUTION: After inputting a text by a node table preparing step 101, nodes, anchors and links to be the constitutional elements of a hypertext are prepared through a sentence dividing step 102, a word decomposing step 103, an anchor extracting step 104, an important position extracting step 105, an anchor marking step 106, and a link preparing step 107. A hypertext display step 108 detemines a node to be displayed in the hypertext and displays a text corresponding to the node. At the time of the display, an anchor is displayed by emphasis such as inversion. When a user selectes the anchor part by a mouse or the like, a node indicating relation with the anchor by a link is displayed instead of the node being displayed at present.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for creating a hypertext structure based on the contents of text, and more particularly to a method for creating, searching and displaying a text database.

[0002]

2. Description of the Related Art Hypertext is electronic text in which a group of texts such as chapters and sections is used as a node and nodes having a content relationship are linked by a link. When the end point of a link is a part within a node, that part is called an anchor.

Salton, G: "Automatic T
ext Processing, "A method based on the word vector space model shown in ADDISON-WESLEY (1989) is known. The appearance distribution of words in a node is represented by a vector, and a combination of nodes with a large inner product of the vector is obtained. In contrast, Shiraishi, et al .: "Online Manual Creation Support Tool," IPSJ 45th National Convention 5C-7, pp.
3-273--3-274 (1992). this is,
A link is created between noun parts appearing more than once in a document, and the importance of the noun part is determined by the particle following the noun part to determine the direction of the link.

[0004]

[Problems to be Solved by the Invention] Salt of the above conventional technique
The method based on the word vector space model of on can be expected to be effective when the entire description contents of a node are related, but as in the manual, a part of the description contents of a node may be related to other nodes. There is a problem that the effect cannot be expected when holding it. Shiraishi and other methods have a problem that the accuracy of link creation cannot be expected. That is,
There is a problem in that noun parts that show subordinate contents instead of the main contents of a node are connected by a link, or a link is created between nodes that have the same noun part but are unrelated to the contents of the node. In addition, it is not possible to create a link about the operation method expressed by a verb.

It is an object of the present invention to refer to certain content so that a reader of hypertext can follow a hypertext reference from a place where he / she wants to know more details to the appropriate place for its description. It is to provide a method of automatically creating a link connecting a location and a node whose content is the main description content.

[0006]

[Means for Solving the Problems] The above-mentioned object is selected from a node table creating step, a sentence dividing step, a word dividing step, an anchor extracting step, an important part extracting step, an anchor marking step, a link creating step, and a hypertext displaying step. It is achieved by the method of making the text hypertext.

Here, the anchor extraction step is composed of a keyword extraction substep and a key phrase extraction substep, and the important part extraction step is composed of an important part temporary recognition substep, an excluded part extraction substep and an extracted part integration substep. The link creation step includes a key phrase type link creation substep and a keyword type link creation substep, and the hypertext display step includes a layout substep, an anchor display position recalculation substep, and a node text display substep.

[0008]

The node table creating step creates a node that is a unit of textual content, and records the connection relationship between nodes and the intended type of node description in the node table. The content grouping of the text is recognized by the data indicating the boundaries of the document constituent elements such as chapters, sections and sections inserted in the input text. The connection relationship between nodes is the order of the nodes in the text and the vertical relationship in chapters, sections, terms, etc. The purpose type of a node's description is the type of description that the node is intended for, such as a brief description or a detailed description, with data indicating the type of description purpose inserted in the text for each node. To do. Document components that do not have data indicating the intended type of description are the same as the document components above them.

In the sentence dividing step, a title and a sentence are cut out from the text in the node by using blanks, punctuation marks and the like as clues. This sentence division technique is a well-known technique related to a machine translation system.

In the word dividing step, the title and sentence cut out in the sentence dividing step are divided into words, and the part of speech and stem of the word and the meaning concept of the word are recognized. This word division technique is also a well-known technique related to a machine translation system.

In the anchor extraction step, in the keyword extraction sub-step, the common noun and compound noun portions are extracted as keywords by comparing the part of speech and character arrangement pattern for keyword extraction with the result of word division. Then, in the key phrase extraction sub-step, by comparing the part of speech and character arrangement pattern for key phrase extraction and the result of word division, the verb or sahen noun and the case element that is semantically connected to the verb or sahenun A pair with an ordinary noun or a compound noun representing the content of is extracted as a key phrase.

In the important point extracting step, the temporary important point extracting step extracts a temporary important point by comparing the pattern for temporary important point extraction with the result of word division. In the excluded part extraction sub-step, the excluded part is extracted by comparing the pattern for extracting the excluded part with the result of the word division. In the extraction point integration sub-step, temporary extraction important points that overlap with the exclusion points are excluded, and the remaining excluded points are regarded as important points.

In the anchor marking step, a flag indicating that it is an important part is attached to a keyword and a key phrase which overlap with the important part.

In the link generation step, in the key phrase type link generation sub-step, the notations, stems and semantic concepts of the verbs or sahen nouns and the nouns or compound nouns constituting the key phrases are the same for the two key phrases. , One keyphrase has a flag of an important point and the other has no flag, and the combination of the target type of the node where each keyphrase exists is allowed, and both nodes are different and direct When there is no hierarchical relationship, a link is created in which a key phrase with no flag of an important point is an anchor of the starting point and a node with a key phrase with a flag of the important point is an end point. The link data includes display range information of the starting point anchor, information of words and phrases forming the starting point anchor, and position information of the end point node.

In the keyword type link creation sub-step, the common nouns and compound nouns extracted as keywords and the common nouns and compound nouns extracted as those constituting the key phrase are regarded as keywords, and 2
For each keyword, create a link according to the same conditions and procedure as the key phrase type link.
That is, in the sub-step of creating a key phrase type link, a link is created by matching a noun or a compound noun, where a link was created by matching a verb or sahen noun and a noun or compound noun.

In both the key phrase type link creation substep and the keyword type link creation substep, the minimum continuous range that includes the key phrase of the anchor that is the starting point or the words that compose the keyword is the anchor display range. However, if there are other links with overlapping anchor display ranges and the end points of the links are different, the minimum consecutive words including the remaining words excluding the overlapping words from the words that make up each key phrase are included. The range is the display range of the anchor at the start point of each link. Also, if there are other links with the same anchor display range and the end points of the links are the same,
Without creating a link, the display range of the anchor that is the starting point of another link is the sum of the display ranges of the anchors.

In the hypertext display step, after the node to be displayed is determined in the layout sub-step, the text of the node is laid out for display, and the document constituent element information and layout information included in the document are deleted. Create a correspondence table of character positions before and after layout. Then, in the anchor display position calculation sub-step, the display position information of the anchor expressed based on the text before the layout is converted into the expression based on the text after the layout by referring to the correspondence table of the character positions before and after the layout. Convert. After that, in the node text display sub-step, the anchor part is displayed by highlighting such as inversion.

As a part of the operation of the link creating step, it has been shown that when the display ranges of the anchors at the starting points of two or more links overlap, the overlap is eliminated, but the elimination processing is not performed. In the node text display sub-step, when a portion where the display range of the anchors overlaps is selected with a mouse or the like to follow the link, highlighting such as inversion of those anchors can be sequentially switched and displayed.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a text hypertext conversion method of the present invention will be described. A flow chart showing an embodiment of this method is shown in FIG. A data flow diagram is shown in FIG.

First, an outline will be described.

FIG. 12 shows an example of text input by this embodiment. After inputting text in the node table creating step 101, a sentence dividing step 102, a word dividing step 103, an anchor extracting step 104,
A node, an anchor, and a link, which are constituent elements of the hypertext, are created through the important point extraction step 105, the anchor marking step 106, and the link creation step 107. An image diagram of the result is shown in FIG. A thin line frame such as a thin line frame 1301 represents a node, a reverse display such as a reverse display portion 1302 represents an anchor, and an arrow 13
Arrows such as 03 represent links.

In the hypertext display step 108, the hypertext display step 108 determines the node to be displayed and displays the text of the node. At that time, the anchor is highlighted and displayed, for example. When the user selects the anchor part with a mouse or the like, the node whose link is indicated by a link is displayed instead of the currently displayed node.

Next, details of the operation of this embodiment will be described.

In the node table creation step 101, text is input to create a node table. A node is a piece of textual content. In this embodiment, nodes are created corresponding to document constituent elements such as chapters, sections, and items. In addition, each node has a purpose of explanation such as “summary explanation” and “detailed explanation”. The node table is a table for recording a range of texts forming a node, a purpose type of a node description, and a child node number list.

FIG. 14 shows a node table created corresponding to the text of FIG. In the figure, "..." of "..." 1408 indicates that characters and items are omitted. The same applies to other charts.

"#" 1401 represents the serial number of the node. “Start” 1404 and “end” 1405 are values that represent the range forming the node by the number of characters from the first character of the input text. "<DOC>" 1201 and "<CHAP>" 120 inserted in the text
2, "<SECT>" 1203, "<SUBSECT
> ”, Such as 1204, to obtain a range of nodes, with data indicating the boundaries of the document components.

The "tag" 1402 is the name of the data indicating the boundary of the document constituent element and represents the type of the document constituent element. “Chapter section number” 1403 is the number of the document constituent element.

The "purpose type" 1406 is the purpose type of the node description. The data indicating the purpose type of the node description is “<NTYPE>” 1205, “</ NTYPE>” 120
It is inserted in the text by the NTYPE tag like 6. So we get the string between this tag as the target type for the node description. A node for which the purpose type of the node description is not shown is assumed to be the same as the node above it.

The "lower node #list" 1407 is a list of nodes below the node. With respect to the node corresponding to the chapter, the nodes corresponding to the sections constituting the chapter are lower nodes. Seen from the node of the section, the node of the chapter becomes the upper node.

"Lower node #list" 1407 in FIG.
FIG. 15 schematically shows the vertical relationship of the nodes indicated by. "#
The number to the left of “#” such as “1” 1501 is the node number and corresponds to the value of “#” 1401.

In the sentence dividing step 102, a title and a sentence are cut out from the text in the node to create a sentence table. In the sentence table, each title, the start position of the sentence, and the number of characters are recorded. An example of the sentence table is shown in FIG. "#" 1601 represents the serial number of the sentence, and "start" 1
Reference numeral 602 is a value in which the start position of the sentence is expressed by the number of characters from the beginning of the text. “Number of characters” 1603 is the number of characters in the cut-out title or sentence, and “sentence” 1604 is a character string of the cut-out title or sentence. A technique for cutting out a title or a sentence from a solid text is a well-known technique related to machine translation.

In the word division step 103, each sentence,
The title is divided into words and the word stem and part of speech are recognized to create a word table. The result of word division of the sentence 1605 is shown in FIG. “#” 1701 represents a serial number of a word in a sentence, and “start” 1702 is a value in which the start position of the word is expressed by the number of characters from the beginning of the text. "word count"
1703 is the number of characters of the word, and “notation” 1704 is the character string of the word. A “word stem” 1705 is a word stem and records a dictionary heading. As for a word that is not registered in the dictionary, its stem is set to "-" as indicated by 1707 "-". “Part of speech” 1706 represents a part of speech. The word division technique is also a well-known technique related to a machine translation system.

In the anchor extraction step 104, an anchor is extracted by associating the result of word division with a previously described anchor extraction pattern to create an anchor table.

FIG. 18 shows an example of the anchor extraction pattern. “#” 1801 represents the serial number of the anchor extraction pattern, and “name” 1802 represents the name of the pattern.
“Content” 1803 represents the content of the pattern. The anchor extraction pattern is described as a sequence of parts of speech and characters by expanding the description method of a character string pattern known as a regular expression.

The anchor table is shown in FIG. "#"
1901 represents the serial number of the anchor. "#" 1902
Represents the number of the node in which the anchor existed, and corresponds to “#” 1401 in the node table. The “purpose type” 1903 represents the purpose type of the description of the node, and corresponds to the “purpose type” 1406 in the node table. “Start” 1904 and “end” 1905 are values that represent the start position and end position of the noun part of the extracted keyword and key phrase by the number of characters from the beginning of the text. The “word stem” 1906 is a combination of the word stems or notations of the nouns or compound nouns of keywords and key phrases. Hereinafter, the noun or compound noun part of the key phrase will be called the noun part. A word registered in the dictionary is a stem, and a word having a stem of "-" is a combination of the notations. "Start" 1907,
“End” 1908 and “word stem” 1909 record the same content as the keyword for the word in the verb or sahen noun part of the key phrase. Hereafter, the verb or sahen noun part of the key phrase is called the predicate part. "Pattern" 1
Reference numeral 910 is the name of the anchor extraction pattern from which the anchor has been extracted, and corresponds to “name” 1802.

The anchor extraction step 104 is composed of a keyword extraction sub-step 301 and a key phrase extraction sub-step 302, as shown in FIG.

In the keyword extraction substep 301, common nouns and compound nouns are extracted according to the keyword extraction pattern. As shown in “W1” 1804, a pattern whose name starts with “W” is a keyword extraction pattern. Keywords are extracted by sequentially applying those patterns.

Here, the description method of the pattern will be briefly described. This description method is an extension of the method for describing a pattern of character sequences, which is known as a regular expression in the field of character string processing, and adds a part of speech description in addition to the character sequence. The part of speech and the character string are distinguished by enclosing the part of speech with "{" and "}" and enclosing the character string with "". Character strings are described by conventional regular expressions. The part-of-speech and the character string are each one sub-pattern, and the sub-patterns can be connected by concatenation or can be connected by AND "&" or OR "|". "(" And ")" represent the range of a group of sub-patterns. "*" Represents 0 or more repetitions of the sub-pattern adjacent to its left side, and "+" represents one or more repetitions of the sub-pattern adjacent to its left side. The extraction range in the pattern is designated by "@". The name of the extraction range is specified by one character following "@". "@@" indicates the end of the range.

The pattern 1805 will be described as an example of the keyword extraction pattern. The pattern 1805 is a pattern for extracting the document name of the manual. In the pattern 1805, the part of speech of the first word is a common noun or an alphabet, the word of which the part of speech is a common noun, a number, or an alphabet is repeated one or more times, followed by one space character, and then by a space character. If there is a character string “grammar / operation manual”, the entire range is extracted with the name “N”. In the keyword extraction sub-step 301, the name "N" of the extraction range is extracted as a portion corresponding to the keyword.

The pattern 1805 allows the anchor 19
As shown in 11, the “DBS2Δ grammar / operation manual” is extracted and recorded in the noun part. Since the predicate part is not specified in the keyword extraction pattern, "-" is filled in each item of the predicate part of the anchor extracted as the keyword in the anchor table.

When there are a plurality of target locations in the extraction range designated by "@" by repeatedly designating "*" and "+", backtracking is performed in the process of associating the pattern with the word division result. Extract all objects. However, if "*" and "+" are inside the extraction range specified by "@" and multiple overlapping ranges are considered as extraction targets, the maximum range is extracted for these overlapping extraction targets. . This means, for example, that a compound noun is a range of one or more repetitions of an ordinary noun, but as a compound noun consisting of three words, one or two parts of it are not extracted and the compound noun is not extracted. Is to extract the entire 3 words.

In the key phrase extraction sub-step 302, a verb or sahen noun and common nouns and compound nouns representing the contents of case elements semantically connected to the verb or sahen noun are extracted by the keyphrase extracting pattern. . As shown in “P1” 1806, the name is “P”
The pattern starting with is the key phrase extraction pattern. Keywords are extracted by sequentially applying those patterns. In the description of the pattern, the name "N" of the extraction range corresponds to the noun part, and "P" corresponds to the predicate part.

When the keyword extracted by the keyword extraction pattern corresponds to the noun part of the key phrase extraction pattern, the keyword is deleted from the anchor table.

As a result of the above processing in the anchor extraction step 104, the anchor table shown in FIG. 19 is created.

In the important point extraction step 105, the important point is extracted by associating the word division result with the previously described important point extraction pattern, and an important point table is created.

FIG. 20 shows an example of an important portion extraction pattern. “#” 2001 represents the serial number of the important part extraction pattern, and “name” 2002 represents the name of the pattern.
"Content" 2003 represents the content of the pattern. Like the anchor extraction pattern, the important point extraction pattern is described as a sequence of parts of speech and characters.

Symbols that were not present when describing the anchor extraction pattern will be described. The "^" at the beginning of the pattern is
Indicates the beginning of a sentence. "[" In the character string indicated by """
And "]" represent a character class, that is, match any one of the characters enclosed in "[" and "]".
When "^" is followed by "^", it indicates that it matches a character other than the negative character class, that is, the characters sandwiched between "^" and "]". "{.}" Represents an arbitrary part of speech.

FIG. 21 shows the important part table. "#"
2101 represents a serial number of an important part. "Start" 210
2, "End" 2103 is a value that represents the start position and end position of the extracted important part by the number of characters from the beginning of the text. The “important part” 2104 is a character string of the part extracted as the important part.

As shown in FIG. 4, the important part extraction step 105 is composed of an important part temporary extraction substep 401, an excluded part extraction substep 402, and an extracted part integration substep 403.

In the important point temporary extraction sub-step 401, a character string is extracted by the important point temporary extraction pattern. As shown in “C1” 2004, a pattern whose name starts with “C” is set as an important portion temporary extraction pattern. The character strings are extracted as temporary important points by sequentially applying those patterns.

As an example of the important portion temporary extraction pattern, the contents of the pattern 2005 will be briefly described. In this pattern, if the number from "0" to "9" or the decimal point "." Is repeated more than one character from the beginning of the sentence and then there is a space character "△", the extraction range name from the next word to the end of the sentence " Take out with "C". That is, the chapter, section, and section names following the number in the chapter, section, and section title lines are extracted.

The character string extracted as the temporary important point is recorded in the temporary important point and excluded point table. FIG. 22 shows a temporary important point and excluded point table. "#" 2201
Indicates the serial numbers of temporary important points and excluded points. “Start” 2202 and “end” 2203 are values that represent the start position and end position of the extracted temporary important part and excluded part by the number of characters from the beginning of the text. “Temporary important part C and excluded part E” 2204 is a character string of the parts extracted as the temporary important part and the excluded part. The “pattern” 2205 is the name of the important part extraction pattern in which the temporary important part and the excluded part are extracted, and the “name” 2
It corresponds to 002.

In FIG. 22, temporary important points 2206 to 2207 represent temporary important points extracted by the important point temporary extraction sub-step.

In the excluded part extraction sub-step 402, a character string is extracted by the excluded part extraction pattern. As shown in “E1” 2006, a pattern whose name starts with “E” is set as an exclusion point extraction pattern. These patterns are sequentially applied to extract a character string as an exclusion point, and record it in a temporary important point and exclusion point table.
In FIG. 22, an excluded part 2208 represents an excluded part extracted by the excluded part extraction sub-step.

In the extraction location integration sub-step 403, the temporary important location where the exclusion location and the range indicated by “start” 2201 and “end” 2202 overlap is deleted. Furthermore, the excluded part is deleted. The remaining temporary important points are recorded in the important point table. The important points are as described above. In this extraction point integration sub-step 403,
Since the temporary important point 2207 and the excluded point 2208 overlap, the temporary important point 2207 is deleted, and the excluded point 22
08 is deleted.

In the anchor marking step 106, an important flag is attached to an anchor whose range overlaps with the important point recorded in the important point table, and an important point recognized anchor table is created. Whether or not the ranges overlap is determined by the “start” 2102 and “end” 210 of the important point table.
3, "start" 1904 of the noun part of the anchor table,
The determination is made by checking the values of “end” 1905, “start” 1907, and “end” 1908 of the predicate.

FIG. 2 shows the anchor table in which the important points are recognized.
3 shows. This is one in which the item of "pattern" 1910 in FIG. 19 of the anchor table is deleted and the item of "important flag" 2310 is added. An anchor having an important flag has a value of 1 in the item of “important flag” 2310, and an anchor having no flag has a value of 0.

In the link creating step 107, a link connecting the anchors in which the stems of the noun part or the predicate part in the anchor table match is created and a link table is created. As a result, the hypertext shown in FIG. 13 is created.

The link table is shown in FIG. "#" 2
401 represents the serial number of the link. The link data includes start point anchor information, end point node information, and stem information.

The structure of the starting point anchor information is as follows. “#” 240, which is a subordinate item of “starting point anchor”
2 is the serial number of the anchor, and corresponds to the serial number “#” 2301 of the anchor table in which the important point has been recognized.
“#” 2403, which is a sub-item of “node”, which is a sub-item of “start point anchor”, is a serial number of a node in which an anchor exists, and a serial number “#” of a sub-item of “node” in the important point recognized anchor table. Corresponding to 2302. The “start” 2404 subordinate item of “node” is a value in which the start position of the node is expressed by the number of characters from the beginning of the text, and the “start” 1404 in the node table of the node represented by the node serial number “#” 2403. It is a copy of the value. "Start" 2405 and "End" 2406, which are subordinate items of "Relative Position", are values in which the display range of the anchor is expressed by the number of characters from the beginning of the node, and "Noun part N" of the important point recognized anchor table.
Sub-items “start” 2304, “end” 2305, and “predicate part P” sub-items “start” 2307 and “end” 230
It is a value obtained by subtracting the value of “start” 2404 of the lower item of “node” from the value of 8.

The structure of the end point node information is as follows. “#” 2407, which is a subordinate item of “end point node”,
“Start” 2408 and “end” 2409 represent the serial number, start position, and end position of the end node of the link. These are “#” 1401 and “start” 140 in the node table.
4, corresponding to “end” 1405. "Start" 2408,
The value of “end” 2409 is expressed by the number of characters from the beginning of the text.

The structure of the word stem information is as follows. The “noun part word stem N” 2410 corresponds to a “word stem” 2306 which is a subordinate item of the “noun part” in the important part recognized anchor table, and the “predicate part stem P” 2411 is “predicate” in the important part recognized anchor table. "Word stem", which is a subordinate item of "Bubu"
Corresponding to 309.

The link creation step 107, as shown in FIG. 5, is a key phrase type link creation substep 50.
1 and a keyword type link creation sub-step 502.

In the key phrase type link creation sub-step 501, a link is created for an anchor of the key phrase type, that is, an anchor in which the "word stem" 2309 of the subordinate item of the "predicate part" of the important point recognized anchor table is not "-". To do. Assuming each of the key phrase type anchors as the starting point anchor, the noun part,
An anchor in which the stems of the predicates match each other is searched, a link with the node at which the anchor is present as an end point is created, and is recorded in the link table. The details are shown in Fig. 6.
Shown in

At the starting point anchor assumption step 601, one of the rows in the important point recognized anchor table as the starting point anchor is set to SA. In the next important flag determination step 602, the important flag of the anchor represented by SA is checked. If the important flag is 1, no link starting from that anchor is created.

If the important flag is 0, in step 603 of searching for a reference destination anchor, the anchor referred to by the SA is searched from the important point recognized anchor table. Here, if the candidate of the reference destination anchor is TA, the TA that meets the following conditions (1) to (4) is set as the reference destination anchor.

(1) SA and TA have the same noun part predicate and predicate part stem.

(2) TA important flag = 1.

(3) A combination of node purpose types of a node with SA and a node with TA is permitted.

(4) The node having SA and the node having TA are different from each other and are not in the direct relation of the upper and lower sides.

In the condition (1), "noun part stem" means
The "word stem" 2306 of the "noun part N" in the important point recognized anchor table is represented, and the "predicate part stem" represents the "word stem" 2309 of the "predicate part P". Similarly, condition (2)
The “important flag” in “” represents the “important flag” 2310. The “node purpose type” in condition (3) represents the “purpose type” 2303. An example of this combination of node purpose types is shown in FIG. "1" 2501 is
Indicates that link creation is permitted, and "0" 2502 is
Indicates that link creation is not allowed. Suppressing unnecessary link creation by limiting the combinations of node purpose types. In condition (4), checking whether or not a node has a direct hierarchical relationship is to check the hierarchical relationship of nodes shown in FIG. 15, and using the value of “#” 2302 as a key, the node table of FIG. "Lower node # list" 1
The determination is made by referring to 407.

If the search is unsuccessful, the link creation process starting from the current SA as the starting point is terminated through the search result determination step 604.

If the search is successful, the display range SAD of the starting point anchor is set to the continuous range from the noun part to the predicate part of SA in the display range setting step 605 of the starting point anchor. That is, “start” 2304 and “end” 230 of the noun part N of the anchor table whose important parts have been recognized.
5, the range between the minimum value and the maximum value of the “start” 2307 and “end” 2308 of the predicate part P is SAD.

At the end point node setting step 606,
The end point node TN is a node where the reference destination anchor TA exists.

Then, in the key phrase type link registration step 607, the display range SA of the starting point anchor is displayed.
After adjusting D, a link having SAD as the starting point and TN as the ending point is registered in the link table. The details of the key phrase type link registration step 607 are shown in FIG.

In the link table search step 701, a link whose starting point display range overlaps with SAD is searched from the link table.

If the search is unsuccessful, it is not necessary to adjust the display range of the starting point anchor, and the SA is selected in the link registration step 703 through the search result determination step 702.
As the information indicating the link between D and TN, the start point anchor information, the end point node information, and the stem are registered in the link table.

When the search in the link table search step 701 is successful, the display range of the starting point anchor is adjusted.

Search result anchor setting step 704
In step SA1, link table search step 70
1 as the starting point anchor of the link of the search result in step 1, and TN1
Is set as the end node of the link of the search result, and TN and TN1 are compared in the node determination step 706.

When TN and TN1 do not match, the noun part of SA and the noun part of SA1 are compared in the noun part determination step 707 of the starting point anchor. The noun part is the “noun part stem N” 2410 in the link table 2.

If the noun parts are different, the SA display range SAD is reset to the SA noun part range in the anchor display range setting step 708. In the subsequent anchor display range changing step 709, the SA1 anchor display range is changed to the noun part range. In changing this, the starting point anchor number of the link table 2402
The key is used as a key to refer to the anchor table in which the important part is recognized, and the range of the noun part is obtained.

This change is made, for example, from the text "change data, table" (data, change),
When two key phrases (table, change) are extracted, and each key phrase serves as a starting point anchor, the range of key phrases overlaps because the verb part is common, so the anchor display range is a noun. It is a part, that is, a "data" or "table" part.

Then, in the link registration step 703, the start point anchor information, the end point node information, and the stem are registered in the link table as the information representing the link between the SAD and the TN.

Step 707 of determining the noun part of the starting point anchor
If the noun part of SA is equal to the noun part of SA1 in S, in anchor display range setting step 710, S
The display range SAD of A is reset to the range of the verb part of SA. In the subsequent anchor display range changing step 711, the display range of the anchor of SA1 is changed to the range of the verb part. Then, in the link registration step 703, S
As the information indicating the link between AD and TN, the start point anchor information, the end point node information, and the stem are registered in the link table.

The processing of changing the display range in steps 710 and 711 is to set the display range of the anchor to "noun part" in step 708 and 709, instead of "verb part". this is,
For example, when two key phrases (data, registration) and (data, deletion) are extracted from the text "data registration / deletion" and each key phrase serves as a starting point anchor, the noun part is common. Since the key phrase ranges overlap with each other, the display range of the anchor is the predicate part, that is, the "register" and "delete" parts.

In the node determination step 706, the TN
And TN1 match, the link is not newly registered, and the display range of the start point anchor SA1 of the link of the search result is the sum of the display range of SA1 and the display range of SAD.

As described above, the processing of FIG. 7 is repeated while assuming that the anchor of each key phrase type of the anchor table in which the important point has been recognized is the starting point anchor SA. As a result, a link 2412, a link 2413, and a link 2414 are created. However, in FIG. 24, the values of the start position 2405 and the end position 2406 of the start anchor of the link 2412 and the link 2413 are not the values obtained as a result of the above-described processing, but are changed by the subsequent keyword type link creation substep 502. The results are shown. This is the end of the description of the key phrase type link creation sub-step 501.

Next, in the keyword type link creation sub-step 502, a link is created for each anchor in the important point recognized anchor table. Regardless of whether "word stem" 2309, which is a subordinate item of "predicate part", is "-" or not, each of the anchors is assumed to be a starting point anchor, and an anchor having the same stem in the noun part is searched for. Create a link that ends at an existing node. The details are shown in FIG.

The flow of FIG. 8 is almost the same as the flow of FIG. The difference between FIG. 6 and FIG. 8 is only the part in FIG. 6 due to the difference between the key phrase and the keyword, and the processing concept is the same. That is, in the flow of FIG. 8, the processing up to the end point node setting step 806 is as follows. A reference destination anchor that matches the starting point anchor SA and the noun part and satisfies other conditions is searched for, and a node having the reference destination anchor is set as TN.

The differences between FIG. 6 and FIG. 8 are as follows. In the reference anchor search step 603 of FIG. 6, the match of the noun part and the predicate part is checked under the condition (1), whereas in the reference anchor search step 803 of FIG. 8, the match of the noun part is checked. Further, in the display range setting step 605 of the starting point anchor shown in FIG. 6, the display range SAD of the starting point anchor is set to be a continuous range from the noun part to the predicate part of SA. At 805, the display range SAD of the starting point anchor is set as the range of the noun part of SA. Further, in FIG. 6, the link is registered in the key phrase type link registration step 607, whereas in FIG. 8, the link is registered in the keyword type link registration step 807.

Keyword type link registration step 8
Details of 07 are shown in FIG.

The processing of the link table search step 901, the search result step 902, and the link registration step 903 is the same as the link table search step 701, the search result step 702, and the link registration step 703 in the key phrase type link registration processing of FIG. Is. That is, when there is no other link in which the display range of the starting point anchor overlaps, the link represented by SAD and TN is registered in the link table as it is. However, even if SA is a key phrase type anchor, the predicate part of the link table is "-" indicating that there is no value.

When the search is successful, the process proceeds to a search result anchor setting step 904 and then to a search result node setting step 905. Node determination step 90
6, the node hierarchical relationship determination step 907, the anchor display range change step 908, and the link registration step 903, the flow proceeds to create a keyword type link and change the display range of the link in which the SA and the start point anchor display range overlap. Processing. The link in which SA and the display range of the starting point anchor overlap is a key phrase type link, and the display range of the starting point anchor of the link of the key phrase type is changed to the range of the predicate part. For example,
The key phrase (data, change) is extracted from the text "Modify data", but the key phrase type link with "Modify data" as the starting point anchor display range and the "Data" starting point anchor display range When creating a keyword type link, the display range of the starting point anchor of the key phrase type link is changed to the range of the predicate “change”.

By this processing, the link 2415 and the link 2416 are created, and the link 2412 and the link 2413 are created.
Change the display range of the starting point anchor. FIG. 24 shows the changed values.

If it is determined that there is no hierarchical relationship in the node hierarchical relationship determination step 907, SAD and TN have no physical relationship and no link is created.

If the TN and the node of the search result match in the node determination step 906, the link is not newly registered, and the display range of the start point anchor SA1 of the search result is set to the display range of SA1 and the display range of the SAD. The sum of

This is the end of the description of the link creation step 107.

By the above processing, the link shown in FIG. 24 is created, and therefore the hypertext shown in FIG. 13 is created. The link 2415 in FIG. 24 is the link 13 in FIG.
03, the link 2412 corresponds to the link 1304, the link 2414 corresponds to the link 1305, the link 2416 corresponds to the link 1306, and the link 2413 corresponds to the link 1307.

Next, hypertext display step 10
At 8, one of the hypertext nodes is displayed. The details are shown in FIG.

In layout step 1001, the text of the node for which the hypertext is displayed is laid out. At this time, the document constituent element information and the layout information included in the document are deleted, and at the same time, the correspondence table of the characters before and after the layout is created. The correspondence table is shown in FIG. “Before layout” 2601 and “after layout” 2602 are
It is a value that represents the position of the character before layout and the position of the character after layout by the number of characters from the beginning of the node.

Anchor display position recalculation step 1002
At, the display position of the anchor is recalculated with reference to the correspondence table shown in FIG. At this time, since the data of the display range of the starting point anchor in the link table is represented by the number of characters from the head of the node, the value of the display range in the link table, that is, the values of “start” 2405 and “end” 2406, respectively. The anchor display position can be recalculated by associating with the value of “before layout” 2601 in the correspondence table and extracting the corresponding value of “after layout” 2602.

Next, the node text display step 100
In 3, the layout result is displayed and the anchor range is highlighted by inversion or the like.

This is the end of the description of the present embodiment.

In the above description, when the links in FIGS. 7 and 9 are registered, if the display ranges of the start point anchors overlap, the overlap is eliminated. Here, as a modification thereof, a method of coping with the hypertext display step will be described, in which it is assumed that the display range of the starting point anchor at the time of registering the link is registered as it is without being resolved.

In the node text display step 1003, when the range of anchors is highlighted and displayed by inversion or the like, only one of the anchors whose ranges overlap is displayed. Then, after the display, the surface of the anchor is switched according to an instruction via a mouse or the like.

FIG. 11 shows the process of switching the anchor's surface when the anchor display portion is selected with the mouse and the mouse button is pressed. In FIG. 11, a mouse input step 1101 of an anchor display portion, a mouse input determination step 1102, an anchor display cancellation step 110
3, link table search step 1104, search result determination step 1105, duplicate anchor display cancellation step 1106, and anchor display step 1107 are the processing for switching the display of anchors for links that include the mouse input location as the starting point anchor range. Is. By this processing, for example, the display of the anchor 2701 in the display shown in FIG. 27 and the anchor 2703 shown in FIG. 28 are switched according to the mouse input. Arrow 2
Reference numeral 702 represents a mouse pointer.

The anchor re-display step 1108 is a processing when there is no other link including the location of the mouse input as the range of the starting point anchor. The processing step 1109 in response to mouse input is processing for performing processing such as link follow-up, instead of switching anchor display.

This is the end of the description of the modified example.

[0109]

By creating a link based on the matching of the stems as the semantic content of a phrase, a link can be created when a part of the description content of a node is related to another node. By using as a key phrase a pair of a verb or sahen noun and a common noun or compound noun that expresses the content of a case element that is semantically connected to the verb or sahen noun, as a key phrase for matching, You can reduce accidental linking. By limiting the combination of the purpose type of the description of the document component of the node where the link start point exists and the end point,
You can reduce inappropriate links.

The display range of the anchor at the start point of the link is set to the minimum continuous range that includes the words and phrases that form the anchor, and the overlapping of the ranges of anchors is eliminated.
Alternatively, by switching and displaying the overlapping anchors, it is possible to provide a user interface in which the user can predict the content of the end point of the link.

Further, by determining the node at the end point of the link based on the result of extracting the important point having the important content within the node based on the expression of each sentence, the content of the anchor at the start point is determined as the end point node. You can select the node that is the main description.

When displaying the hypertext, by recalculating the display position of the anchor based on the correspondence between the characters before and after the layout, the link data before the layout is commonly used even when different layout rules are applied. be able to.

[Brief description of drawings]

FIG. 1 is a flow chart showing an embodiment of a text hypertext conversion method according to the present invention.

FIG. 2 is a data flow diagram showing an embodiment of a text hypertext conversion method of the present invention.

FIG. 3 is a flowchart showing the contents of an anchor extraction sub-step in the text hypertext conversion method.

FIG. 4 is a flowchart showing the contents of an important part extraction sub-step in the text hypertext conversion method.

FIG. 5 is a flowchart showing the contents of a link creation substep in the text hypertext conversion method.

FIG. 6 is a flowchart showing the contents of a key phrase type link creation step in the link creation substep.

FIG. 7 is a flowchart showing the contents of a key phrase type link registration step in a key phrase type link creation step.

FIG. 8 is a flowchart showing the contents of a keyword type link creation step in the link creation substep.

FIG. 9 is a flowchart showing the contents of a keyword type link registration step in the keyword type link creation step.

FIG. 10 is a flowchart showing the contents of a hypertext display sub-step in the text hypertext conversion method.

FIG. 11 is a flowchart showing anchor display switching processing in a hypertext display sub-step.

FIG. 12 is a diagram showing an example of a text input to the hypertext conversion method for text of the present invention.

FIG. 13 is a diagram showing an example of hypertext created by the text hypertext conversion method of the present invention.

FIG. 14 is a diagram illustrating an example of a node table that records a result of creating a node from text.

FIG. 15 is a diagram schematically showing the vertical relationship of nodes in a node table.

FIG. 16 is a diagram illustrating an example of a sentence table that records a result of cutting out a sentence in a text.

FIG. 17 is a diagram illustrating an example of a word table that records a result of dividing a sentence in a text into words.

FIG. 18 is a diagram illustrating an example of a pattern for extracting an anchor.

FIG. 19 is a diagram illustrating an example of an anchor table that records extracted anchors.

FIG. 20 is a diagram illustrating an example of a pattern for extracting an important part.

FIG. 21 is a diagram showing an example of an important point table recording the extracted important points.

FIG. 22 is a diagram showing an example of a temporary important point and excluded point table created on the assumption that important points are extracted.

FIG. 23 is a diagram showing an example of an important-parts-recognized anchor table as a result of attaching an important-part flag to each anchor of the anchor table based on the contents of the important-part table.

FIG. 24 is a diagram illustrating an example of a link table that records created links.

FIG. 25 is a diagram showing a node purpose type matrix that specifies a combination of purpose types for explaining nodes at a start point and an end point of a link when a link is created.

FIG. 26 is a diagram showing an example of a correspondence table of character positions before and after layout, which is created in the layout sub-step in the hypertext display step.

FIG. 27 is a diagram illustrating an example of a display state of anchors.

[Explanation of symbols]

101 ... Node table creation step, 102 ... Sentence division step, 103 ... Word division step, 104 ... Anchor extraction step, 105 ... Important point extraction step, 10
6 ... Anchor marking step, 107 ... Link creation step, 108 ... Hypertext display step.

Claims (8)

[Claims] 1. Inputting text in which data indicating the boundaries of document constituent elements such as chapters, sections, and terms is input, and creating a node that is a semantic unit of text based on the data, and from each sentence Extract the words and phrases, extract the important points that have important contents in the node, flag the words that overlap with the important points as important points, and use the words that do not have the important points flag as the anchor of the starting point In the hypertext conversion method of text that converts text to hypertext by creating a link that indicates the relationship between the two, with a node that has a phrase that matches all or part of the word having an important point flag as the end point. The words and phrases extracted from the sentence consist of the keywords consisting of common nouns and compound nouns from each sentence in the node, and words and phrases that have a semantic connection. And to extract and key phrases,
Extracting important points that have important contents in a node based on the expression of each sentence, and dividing words into words and checking each word's notation or A method for converting text into hypertext, characterized by comparing stems or semantic concepts. 2. A method for extracting a key phrase, which comprises dividing a sentence into words, recognizing a part of speech, and extracting a part where the part of speech of a word and the arrangement of characters match a predetermined pattern as a key phrase. The method for converting text into a hypertext according to claim 1. 3. A verb or sahen noun and a common noun or compound noun representing the content of a case element that is semantically connected to the verb or sahen noun are defined as words in a semantic connection that form a key phrase. The hypertext conversion method according to claim 1, wherein the hypertext conversion is performed. 4. In addition to data indicating boundaries of document components such as chapters, sections, and paragraphs, data indicating the intended type of description of document components is inserted as input text, and links indicating relationships are input. 2. When creating, the link is created only when the combination of the target type of the node having the link start point and the node end point of the link is a fixed combination. Hypertext conversion method 5. The text according to claim 1, wherein when the link is created, a link is created in which a minimum range of consecutive words including a word constituting the anchor of the starting point is set as the display range of the anchor. Hypertext conversion method. 6. When a part within the range of the anchor is selected with a mouse or the like to follow the link, there are two links including the selected part as the display range of the anchor of the starting point.
When there are more than one, a hypertext display method characterized by displaying the display range of those anchors sequentially by highlighting such as reversing each time a selection operation with a mouse or the like is performed. 7. When creating a link, a temporary minimum display range of anchors that includes the words and phrases forming the starting anchor is used as a temporary display range of the anchor, and a link starting from another anchor having the overlapping range is set. In some cases, the display range of each anchor is set to a minimum continuous range including words that do not overlap in terms of words forming the anchor, and the hypertext conversion method of the text according to claim 1. 8. When displaying the text by deleting the document component information and the layout information included in the text by laying out the text, the text before the layout and the characters of the text after the layout are associated with each other, and Hypertext characterized by displaying the anchor after converting the display range information of the anchor at the start point of the link expressed based on the position of the character to the display range information based on the position of the character of the text after layout. Display method.