JPS58225485A - Automatic production system for book index - Google Patents

Abstract

PURPOSE:To produce a high-reliability, high-level index page by adding level discrimination flag corresponding to respective decomposed terms, and utilizing information such as the level discrimination flag for thinning out unnecessary term on specified basis. CONSTITUTION:A storage area 8 for a text is read and a control code extraction part 9 extracts one control code and passes processing to a preprocessing part 10. The preprocessing part 10 analyzes one index data inputted together with a separation code to check the number of separated elements and characteristics of the element. An arithmetic processing part 12 is give with the separated elements corresponding discrimination flag. A postprocessing part 13 thins out unnecessary element among combination of the separated elements on the specified basis by utilizing the level discrimination flag. Data from the postprocessing part 13 is outputted to an output processing part 15 through an index record file 14.

Description

Detailed Description of the Invention (a1 Technical Field of the Invention The present invention relates to a method for automatically generating an index provided at the end of a book (BL technology) in a document processing system in which the layout of a book is automatically performed by a computer. Background FIG. 1 is a block diagram of a document processing system. First, a key buncher inputs a manuscript l of the main text of a book using an input device 2, and stores it on a magnetic medium 3.

By inputting this information into the computer system and specifying and inputting layout information such as the number of lines to be printed per page, the editing section 4 automatically allocates it in the computer's memory and prints it on a kanji line printer, etc. If you output it with , you will have a version 5 of the main text. Also, the index processing unit 6
Then, terms that will serve as index headings are automatically extracted from the main text storage area 8, arranged in alphabetical order, for example, and the index page 7 is printed out.

(C) Prior art and its problems In order to automatically generate an index in this way, conventionally, if the term "orthogonal transformation of a real symmetric matrix" is used in the text, it must be added to the index. If desired, mark the terms in the text with control codes, etc., and also indicate how to read them. For example, (IX) Orthogonal transformation of a real symmetric matrix (K) Jicky Shogyoretsu no Chokkou Henkan (IE)
Enter it in the text.

Here, (IX), (K), and (IB) are control codes, respectively. (IX): Start of index (K): Reading and delimitation (IE): End of index. Read out the area and check it from beginning to end, extract terms that have control codes like the ones above, classify them in alphabetical order, add the number of pages, and create a C index page. .

However, if we simply use the terms in the main text as headwords in the index, the 1111i1
It is inconvenient for readers to search because they can only be indexed from the beginning, and the headwords for the index are limited. In other words, since indexes exist to facilitate searches by readers, it is desirable that one term be classified from various viewpoints. Particularly in the case of academic books, it is useful to have a rich index structure that allows you to break down one term and index each broken down term, allowing readers to easily and reliably find the desired term. You can see the difference.

・For example, the term “orthogonal transformation of a real symmetric matrix” mentioned above can of course be used as an index entry as before, but in addition, as shown in Figure 2, “
The headword ``transformation'' is also provided, and accompanying headwords such as ``orthogonal'', ``Fourier'', and unitary are provided at lower levels. Here, “conversion” is inserted for non-blue.

Further, accompanying the headword "orthogonal", there are also lower-level terms such as "-" for a symmetric matrix and "- for a symmetric operator."

Here, "orthogonal transformation" is inserted into non-blue.

In this way, the terms in the main text are broken down, and the broken down terms and “
In combination with non-blue or more related terms denoted by ``-'', ``transform'' is at the first level, ``orthogonal'' and ``Fourier'' are at the second level, ``-'' for a symmetric matrix and ``for a symmetric operator -" is the third level, resulting in a fine-grained (and easy-to-use) index structure.

In addition to "transformation" in FIG. 2, it is preferable to use a headword such as "orthogonal" for "symmetrical" matrices to be indexed.

In this way, an index that is broken down into each word and has subordination relationships and mutual reference relationships, and is divided into levels for each hierarchy is most preferable and is indispensable for books with high-quality content. By the way, such advanced indexes have traditionally been created manually, but the method of breaking down headwords and making connections is difficult.
Index Editor S Person 5 - J - It is a problem that is difficult to deal with even in computer-based document processing systems, as it depends largely on knowledge and skills.

(d1 Purpose of the Invention The purpose of the present invention is to solve such problems in the conventional automatic index generation system, and to break down the terms in the text and create a rich index that covers related terms without any manual intervention. The purpose is to be able to create it automatically.

(e) Structure of the Invention In order to achieve this object, the present invention involves inputting an index code by adding an index code to the terms used in the heading of the index when inputting a manuscript of the main text, and recording the code in a storage means. In the automatic generation method using a computer system for book indexes, which extracts and categorizes index terms and page numbers by sequentially checking the data of the text after completing the layout of the main text, in the case of compound words, they are Decompose it into each element, add a separator to the joint between each element, input it, and based on that, give a level identification flag corresponding to each decomposed term in the computer, read it out, and input it. In addition to creating all combinations of terms that have been decomposed, it also generates lower-level terms that accompany each decomposed term, and uses information such as level identification flags among the combinations of terms that have been decomposed as described above. A configuration is adopted in which post-processing is performed to thin out unnecessary terms based on certain criteria.

ff1 Embodiments of the Invention Next, how the automatic book index generation method according to the present invention is actually implemented will be explained by way of embodiments. FIG. 3 is a block diagram showing a processing method according to the present invention. In the case of the present invention, the main text storage area 8 is read out, the control code is extracted, and an index page is generated, as in the case of the conventional method, but a separator mark for compound words is added as follows. .

In other words, in order to be able to decompose the terms in the main text and extract the smallest unit of terms, when entering the term in the main text, if the term is "orthogonal transformation of a real symmetric matrix", as in Example 1, Disassemble as much as possible and add separators.

Example 1... (IX) real (Ill symmetric (p1 matrix (
+) l Orthogonal (p) transformation (k) Jitsu (p) Taisho (p
) Goretsuno (P) Chosokou (p) Henkan (IE
)... If you enter it like this, the computer will automatically process it after editing the main text. That is, the main text storage area 8 is read out, one control code is extracted by the control code extraction unit 9, and the processing is passed to the next preprocessing unit 10. Then, the preprocessing section l
In O, one index data input with a separator is analyzed, and the number of separated elements and characteristics of the elements are checked. 1
1 indicates the storage state of one term after preprocessing, and is stored in the form of each decomposed index term, its pronunciation, and a separator flag. In this example, the number of elements is 5 because it is decomposed into 5 parts as described above.

Some of the decomposed terms are handled differently. Most of the words used in the index are in the form of "adnominal modifiers", and separators include those between the adnominal modifier and the modified word, and those between each compound element in a compound noun. There is something.

The former will be referred to as A level and the latter as B level. For example, in the compound noun "real symmetric matrix,""real,""symmetric," and "matrix" each have a high degree of closeness, so the space between them is set to B level, and the modifier "of a real symmetric matrix" is used.
Since there is a low degree of closeness between this term and the modified word "orthogonal transformation," it is set at A level. In the illustrated example, A
Other words are considered as B level, and the level is determined based on the presence or absence of kana at the end of the word and stored in the separator flag.

Note that this separator is used in the thinning process described later.

Further, each element term is assigned a number such as 0 to 4 and stored, so that it can be handled by replacing it with a number from now on.

Therefore, the index structure as shown in FIG. 2 is a permutation obtained by arranging the numbers corresponding to each element in order from level 1 in the storage area of the computer, that is, 9-.

Moreover, this permutation is obtained by using a classification program in the calculation processing section 12 and performing calculation processing according to the processing flow shown in FIG. In this figure, the functions of each processing block are as follows. That is, INTL is X = (
0),'f(1)=(0). , CIIIRL repeatedly executes the process up to the corresponding terminal ■ for each value while changing the value of the control variable μ one by one from 1 to m.

LOAD retrieves f (j) from memory. 5H
FT converts each character of the elements of list f (j) into i−j+
l shift. C0N5 concatenates list X and list y to obtain list (x, y). EXTN indicates that the elements of list Z are characters 0, 1 . ..., expand so that it becomes a permutation of i. That is, for the elements of Z, 0, 1 . ...,
If there are any characters not included in i, their ascending sequence is added after that element. 5TOR stores the contents of the variable 10- into f (i+l) in memory.

Through such list processing, in the case of Example 1, if the number of elements is found as n=5 and all combinations are generated by permuting the numbers, 34 combinations can be created as shown in Table 1. In other words, the column with item number 1 has exactly the same term as the term in the original text, and is placed in the first hierarchy. In this case, of course, there are no layers below the second level. Item numbers 2 to 5 are replaced, the level of the hierarchy is "symmetrical", the second level is "actual", and the third hierarchy does not exist. From item number 6 onwards, the replacements become progressively more complex, and as seen in some of the items marked with an ○, there are cases where there are more than 4th hierarchy.

However, it is not necessary to use all such combinations as headwords in the index. For example, if similar words appear in a concentrated manner on the index page of a book, it is not only unsightly, but also due to space constraints, and may not be necessary for index searches. Therefore, the hierarchical level checking section 131 and headword concatenation checking section 132 in the post-processing section 13 consider that items that do not meet the following conditions are inappropriate and do not output them to the index page.

Condition 1: When the number of hierarchy levels exceeds the standard (usually 2 to 3).

Condition 2: When the intermediate result of successively concatenating the level or the lower hierarchy includes an A-level separator, and the end of the concatenated next level is a B-level separator. For example, “matrix (ρ) orthogonal (pi”).

Based on these conditions, excluding the applicable ones, only the combinations marked with an O in Table 1 are valid.
An index record file 14 is created. In Table 1, the combinations that are marked "Yes" in the "4~" column are those that are omitted by the hierarchy level check unit 131 because the number of hierarchy levels in condition 1 is too large, and other combinations are is condition 2
This is a combination that is omitted by the headword concatenation check unit 132 for semantic reasons. For example, combinations such as item numbers 4 and 8 are removed because the separator after "matrix" is an A-level separator, and the separator after "orthogonal" is a B-level separator, and satisfies condition 2.

13- When this post-processing is completed, the index term extraction unit 9 extracts the next index term again, and the pre-processing-calculation process-post-processing is repeated.

If the number of index words is still too large even after being thinned out in the post-processing unit 13, groups are thinned out at the output processing stage. That is, the index records are classified and accumulated based on the post-processed index terms, and the index record file 14 is created and stored. As shown in the figure, this index record is made up of identification numbers representing headwords, readings, nonples, and derived words at 1llll'it level.

Figure 5 shows an enlarged view of a portion of the index record 14.As shown in this figure, the reading method is to inherit the items in the upper hierarchy in order for the classification to be successful. I'll try to go with it. This is then allocated to an index page like in a conventional document processing system.

Next, the assigned results are re-inputted, and the output processing unit 15 selects headwords derived from the same word (judged by identification number) at a certain hierarchical level in the alphabetical order or alphabetical list. ) is divided within a certain distance (=j, for example within 10), one of them will be deleted according to the following rules.In other words, the purpose of the index is to facilitate the reader's search. Therefore, it is desirable for one term to be classified from various viewpoints.Therefore, it does not make much sense even if the same common words are concentrated in the same vicinity.Especially for books that do not have enough space for index pages due to space constraints. In some cases, there is a desire to keep headwords to the minimum necessary.

The rules for deleting such terms that have little function as headwords are: When the depth of the first layer of engineering differs, the deeper one is more complicated, so it is deleted.

2. If the depth of the hierarchy is the same, delete according to the following rule.

(a) If a non-blue word is sandwiched between headwords from both sides, such as "one transformation of a matrix," it will be difficult to see it if it is located at a higher hierarchical level. Sharpen. If they are at the same level, remove the one with more layers that apply to this rule.

(If the rule (a) on BL does not reduce the amount, delete the layer with the most A-level separators disabled.

FIG. 6 shows an example in which the ``see-through'' process is performed based on this rule, where ``■■■■'' is too close to the second level, but ``■'' cannot be deleted because it has a headword with a different identification symbol in a lower level. Also, in ■■■, rule 1. Accordingly, ■ that does not have a third layer is left.

As a result, the index page 16 that is finally output will be in a state as shown in Example 2.

Example 2 Orthogonality of transformed real symmetric matrices - Similarity - Orthogonality - Orthogonality of real symmetric matrices - 1C one-plane figure - Fourier seek. Note that the index that has been subjected to the see-through processing is stored and saved in the checked index page file 17.

(g1 Effect of the Invention According to the present invention, when an index term is a compound word, if it is broken down into each element and a separator is added to the joint between each element and input, the computer calculates the , a level identification flag corresponding to each decomposed term is given, and it is read out to create all combinations of decomposed terms, as well as to assign an index to each decomposed term.
It also generates lower terms in the associated hierarchy, and uses information such as level identification flags among the combinations of terms decomposed as described above to eliminate unnecessary terms based on certain criteria. The system employs a configuration that performs thinning (post-processing).

Therefore, unlike the conventional case of using the terms in the text as they are as headwords in the index, all combinations of terms are generated by decomposing the terms in the text, and only the necessary headwords are retained. become. As a result, necessary and sufficient headwords can be obtained, and since they are realized through computer processing, problems caused by the editor's discretion are eliminated, making it possible to create highly reliable index pages. .

[Brief Description of the Drawings] Figure 1 is a block diagram showing the general configuration of a document processing system, Figure 2 is an example of an index with hierarchical levels, Figure 3 is a block diagram showing a processing method according to the present invention, FIG. 4 is a block diagram of the calculation processing section, FIG. 5 is an enlarged view of a part of an index record, and FIG. 6 is a diagram showing an example of a ``seeing''. In the figure, 6 is an index processing unit, 9 is an index term extraction unit,
10 is a preprocessing section, 11 is a storage 18-state after preprocessing, 12 is a calculation processing section, 13 is a postprocessing section, 14 is an index record file, and 15 is an output processing section. Patent Applicant: Fujitsu Limited Agent, Patent Attorney Minoru AoyagiGo to Figure 1, Figure 1, and Go to Main Text. Commissioner of the Japan Patent Office Kazuo Wakasugi 1, Indication of the case: Japanese Patent Application No. 57-1094982, Title of the invention: Automatic book index generation method % formula % 5, Subject of amendment: "Detailed description of the invention" column 6 of the specification , Contents of the amendment As shown in the attached sheet 1. "Reading and delimitation" on page 3, line 15 of the specification has been changed to "
It is corrected as "separation from reading." 2. "Non-pull" on page 4, line 19, page 5, lines 3 and 6, and page 15, line 18, are corrected to "dash". 3. On page 9, line 9, "terminology" is amended to "terminology". 4. Correct "classification program" in lines 5 and 6 of page 10 to "replacement processing program." 5. Correct "from l to m" in line 11 of page 10 to "from 1 to input value m". 6. r LOAD on page 10, lines 13 to 18 is as follows:
~Expanding. ” shall be amended as follows. rLo^D is the I N corresponding to the specified number j)
retrieve from memory. S II F T shifts each character of the elements of list f (j) by i-j+l using the specified number i. C0N5 concatenates the two input lists, list X and list y, to create list (x +
y) is obtained. EXTN is the input or the element of the strike z,
Based on the specified number l, the characters 0.1. .... Expand so that it becomes a permutation of i. 7. In the first line of page 11, “in memory” is changed to “specified i”.
"In memory" according to the correction. 8. On page 11, line 11 of the same page, change “O-marked Tachino” to “O
"Things with no markings," he corrected. 9. Correct the "identification number" on page 14, line 11 and page 15, line 2, respectively, to "identification symbol." 10. Patent applicant characterized by "orthogonal" unpublished on page 16 of the same Patent attorney Minoru Aoyagi, patent attorney 1. Indication of case Patent application 1984-1094982, Invention Name: Book index automatic generation method 5, date of correction order: September 9, 1980 ■0 Figures 3, 4, and 6 of the drawings will be corrected as shown in the attached sheet. 2, page 16, lines 7 to 8 of the specification [Figure 6 shows
In this example, the statement "is amended as follows. ``Figure 6 (a) shows an example of ``see-you-ku'' processing based on this rule, and (b) shows the correspondence between ``index terms in the text'' and ``identification symbols'' in that case. This is a diagram. ”

Claims (1)

[Claims] When inputting the manuscript of the main text, an index code is added to the term used in the index heading, and is recorded in the storage means. In the automatic generation method using a computer system for book indexes, which extracts and categorizes index terms and page numbers by sequentially checking the data, in the case of a compound word, it is broken down into each element and the joint between each element is A separator is added and input, and based on that, a level identification flag corresponding to each decomposed term is given in the calculator, and it is read out to create all combinations of decomposed terms, and each It also generates lower-level terms that accompany the decomposed terms, and uses information such as level identification flags to thin out unnecessary terms based on certain criteria from among the combinations of each decomposed term as described above. An automatic book index generation method characterized by processing.