JPH01114971A - Device for supporting document formation/calibration - Google Patents

Abstract

PURPOSE:To simplify decision by individually displaying measured data based upon relative relation between a reference value of a parameter and its measured value in case of evaluating a document. CONSTITUTION:When an evaluated sentence stored in an input character storing means 2 is read out to a display device and language analysis processing is executed, the evaluated sentence is analyzed, a position to be calibrated is extracted and calibration analysis processing is executed. The measured values of respective parameters indicating readability are found out from the processed results of format element analysis and calibration analysis and stored in a buffer of a control means 5, the evaluated value is expressed by a bar graph and the evaluated value itself is displayed. The parameters are divided into three branches by comparing the measured value of each evaluation parameter with a recommended value and respective branches and individually displayed.

Description

[Detailed Description of the Invention] <Industrial Application Field> The present invention is useful for extracting errors and inappropriate expressions from Japanese texts, providing guidance and correction information, and necessary! ! The present invention relates to a document creation/proofreading support device that creates documents by providing information such as information and other information. It can also be used as a speech-to-speech device that performs language analysis on sentences containing kana and kanji, obtains kana character strings, and performs speech synthesis using the results.

<Conventional technology> Currently, Japanese word processors (hereinafter referred to as word processors) are in practical use, and related basic technologies such as Japanese input/output, editing, kana-kanji conversion algorithms, and dictionary technology have been established. are doing.

Word processors that use the kana-kanji conversion method require a dictionary for conversion, and user dictionary technology has been established that allows users to add their own words.

Furthermore, basic techniques such as morphological analysis, syntactic analysis, and semantic analysis are known in Japanese language processing technology.

In Europe and the United States, word processors became popular early on, so related technology has advanced, and devices with spell check and correct functions have been put into practical use.

In addition, although it is currently in the development or prototype stage, the device uses morphological information to diagnose the readability of sentences, and aims to support users in writing easy-to-understand sentences by notifying the user. The device has also been announced.

Compared to Western words, which are written with punctuation into each word, Japanese words are usually written in kanji, kana, and kanji letters without punctuation, and the orthography is not very strict. In addition, analysis is difficult and no equipment for automating calibration has been put into practical use.

Traditionally, when accurate Japanese was required, a pair of translators would work together to read the text and extract problematic parts, or someone with specialized proofreading knowledge would cross-check the text one by one. Methods such as adding and calibrating are used.

Recently, devices for supporting such calibration work are being developed.

Above, we have described conventional technologies centered on language processing, but in addition to these technologies, technologies related to workstations have been established, and devices using multi-windows to improve work efficiency have been put into practical use. ing.

<Problems to be Solved by the Invention> In recent years, Japanese word processors have become widespread, and an increasing number of documents are created using these devices. Many word processors are now using the kana-kanji conversion method (hereinafter, unless otherwise specified, this includes the romaji-kanji conversion method), which makes input easier.

Kana-Kanji conversion algorithms generally use word dictionaries, linguistic information such as grammar, and probabilistic information such as word frequency in the process of converting kana to kanji.
Correct conversion results are not always output.

For this reason, users of word processors, etc., may not notice incorrect conversion results, may forget to correct the errors even after noticing them, or may make a single human error, resulting in incorrect sentences. It may occur that you create one.

A document creation/proofreading support device extracts such mistakes, presents correction information, and automatically corrects them.

One of the functions of this device is a method for improving sentences by expressing the readability of sentences using morpheme information in numerical or relative values and presenting the results to the device user. It is machined. A currently known method is to present readability parameters, measured values of the parameters in evaluation sentences, and an overall readability index.

However, conventional methods only display numerical values,
The drawback was that it was difficult to understand the meaning of the numbers.

Another drawback is that it is difficult to determine whether a parameter value exceeds, falls short of, or is appropriate in the evaluation text.

In addition, there was a drawback that it was difficult to understand which items had problems in relation to the entire text.

Furthermore, since problems are not clearly displayed, there is a drawback that even if the text is evaluated, it is not possible to obtain guidelines as to what parts should be corrected.

The present invention attempts to solve this problem by visualizing the values of text evaluation parameters using a graph or the like and displaying them separately according to the degree of evaluation.

Means for Solving Problems> The present invention provides a means for inputting and editing Japanese, a means for storing the input Japanese theory, a means for storing a dictionary, a means for storing grammar, A control means such as a microprocessor that converts and edits the input kana character string into a sentence mixed with kanji, a means for displaying a character/symbol string, and a character/symbol string that needs to be proofread. It consists of means for modifying the character string.

<Operation> The present invention operates so that when a sentence to be evaluated is analyzed using a character code or a language analysis function and the results are displayed, the evaluation parameters and their actual measured values are displayed together with a graph for easy viewing.

Furthermore, by dividing the graph by comparing it with the recommended value of the evaluation parameter, it is possible to present the intermediate score in the sentence in an easy-to-understand manner to the user.

<Example> The present invention will be described in detail below based on the drawings. FIG. 1 is a block diagram of a document creation/proofreading support device according to the present invention.

In the figure, l is a means such as a keyboard for inputting and editing Japanese character strings. This includes a kana-kanji conversion function that turns kana into kanji, which is now a well-known fact.
It also includes the ability to specify a certain string.

2 is a means for storing the Japanese character string inputted by the input means. A keyboard is usually used as the input means, but it is also possible to use an external storage means such as a floppy disk or magnetic tape for storing input Japanese character strings without sequential input. That is, there may also be a configuration in which the l input means is omitted.

3 is a means for storing a dictionary for analyzing the Japanese character/symbol strings accumulated in 2 above, and this also includes a means for storing a user dictionary that can be defined, registered, and deleted by the user. .

4 is a means for storing rules for analyzing grammar and other sentences.

5 extracts a part of the string stored in 2,
This is a control means that stores intermediate results and commands display. The control means includes means for storing results obtained by the control.

Reference numeral 6 denotes display means such as a CRT for displaying input character strings, intermediate results of verification, character strings to be proofread, KWIC (Keyword In Context), and the like.

Reference numeral 7 denotes a proofreading means for correctly reflecting the results of corrections made to the portion to be proofread indicated by 6 in the original text.

In order to provide a concrete explanation, a case will be described below in which the readability of a text is evaluated. However, the present invention is not limited to evaluating the readability of sentences, but can be broadly applied to various evaluations for which reference values are set.

FIG. 2 is a diagram showing the results of analyzing the calibration displayed on the display means of the apparatus. 8 is a sentence to be evaluated for readability, and the shaded portion in the figure indicates that the device has extracted a location where there is a possibility of an error.

The figure shows a typical example for explanation. 9 is an example of a typographical error, which is an example of a homonym error caused by a defect in the kana-kanji conversion software. 10 is an example of a mistake in the same sound kana, 11 is an example of a mistake in okurikana, 12 is an example of an omission, and 13 is an example of insertion of an unnecessary character.

FIG. 3 is a diagram showing an example of the structure of a buffer that stores parameters representing readability. Reference numeral 14 contains a parameter representing the readability of the text, 15 contains the measured value of each parameter of the text to be evaluated, 16 contains the recommended value of each parameter, and 17 contains the evaluation value of each parameter.

Among the parameters in FIG. 3, 18 represents morpheme information,
19 represents calibration information. As an example of No. 18, in FIG. 3, the proportion of Chinese characters included, the average sentence length, and the average frequency of punctuation marks are increased, and as an example of No. 19, the ratio of occurrences of proofread clauses is increased.

Here, the parameter measurement value 15 in FIG. 3 will be specifically explained. The total number of characters in the evaluation target sentence 8 is 45 including punctuation marks. Although there is a method to calculate the total number of characters excluding punctuation marks, you can calculate the value using a method similar to the method described below. There are 20 kanji in 8, and the percentage of kanji included is 20/45=0.444, which is 44.4%, if you calculate it to the second decimal place. The average sentence length is 45/2 = 22.5, and the average punctuation frequency is 45/2 = 22°5.
The proof clause appearance ratio = 5/14 = 0.36.

Next, the recommended values will be explained. Although the recommended value differs slightly depending on the individual, it is often determined empirically as a common sense value for each ethnic group. This is similar to how a language system is created sequentially as a group of people's will, rather than first determining the grammar and terminology for the words we use within the membrane and then proceeding with the actual use of the language. ing.

Such values include the recommended values of kanji ratio, average sentence length, average punctuation frequency, and number of proofreading extractions, RISR2,
When R3 and R4 are 15, we obtain 25≦R1≦45, 25≦R2≦45 7≦R3≦15 R4≦0.1. However, the present invention does not refer to the absolute value of this value itself.

Next, the evaluation value is obtained by evaluating the difference between the recommended value and the actual value of the sentence to be evaluated using some evaluation function.

To simplify the explanation, an extreme example will be given here in which the value is 100 if the actual measured value is the recommended value, and 0 otherwise. 17 in FIG. 3 is the value obtained in this way.

Another example of how to calculate the evaluation value is trapezoidal approximation. This is a method in which the evaluation value is set to 100 between designated areas, and linear approximation is applied to areas that deviate from the designated values.

Naturally, there is also a method of assigning a nonlinear function to the deviant portion. Since the present invention does not limit the approximation function itself, it will only be mentioned that various approximation functions can be applied.

FIG. 4 is a diagram showing the results of evaluating the readability of sentences according to the present invention.

The vertical axis shows the evaluation parameters, and the horizontal axis shows the relative values of the parameters. 20 represents a specific example of a parameter, and 21 represents a value of the parameter.

22 is a bar graph corresponding to the value of 21. The parameter values are normalized so that the recommended value for each parameter ranges from 0 to 100.

In actual display, the parameter value is designed to indicate 100 if it exceeds 100, and to indicate 0 if it is less than or equal to 0.

Figure 4 shows the readability evaluation results using a bar graph.
The present invention is not affected even if the graph is not a bar graph. for example,
Another example of a representation is a radar chart.

Now that the above preparations have been made, the core part of the present invention will be described.

FIG. 5(a) is a diagram in which only the portion of the content ratio of kanji characters in FIG. 4 is extracted. Moreover, FIGS. 5(b) and 5(C) are diagrams corresponding to when different evaluation sentences are evaluated and the value is 50% and 10%, respectively. To avoid confusion, number the lengths of the bar graphs and numbers. 2
3 is the length of the bar graph of the proportion of kanji characters included when evaluating the evaluation sentence 8, and 24 is a numerical value. 25 is the length of a bar graph when another sentence containing 50 kanji characters is evaluated, and 26 is its value. 27 has a kanji ratio of l
It is the length of the bar graph when evaluating another sentence of O,
28 is its value.

One of the features of the present invention is that the readability parameters of sentences are shown in bar graphs and actual values to make them easier to visually see.

Figures 6(a), (b), and (C) are respectively shown in Figure 5(a).
), (b), and (C), and bar graphs are displayed differently according to their values. 29 is the length of the Keyaki graph for sentence 8 with a kanji content ratio of 44.4%, and 30 is its value. 31 is the length of a bar graph of a sentence containing 50% kanji, and 32 is its value. 3
3 is the length of a bar graph of a sentence containing 10% of kanji characters, and 34 is its value.

Here, a description will be given of how such a display is realized. The recommended value R1 of the ratio of kanji included in the example given now is expressed as 25≦rtt≦45.

By comparing the rtt and the percentage of kanji included in the evaluated text, it can be classified into the following three types. That is, the included kanji comparison is greater than, equal to, or less than the recommended value.

If these three types of values are distinguished and displayed, it becomes easier to determine what state the parameter is in with respect to the recommended value.

Figure 6 shows the proportion of kanji included is higher than the recommended value. If it is small, use a seal If they are equal, The larger value indicates the case where 4HHH is used.

In FIG. 6, the distinction information is displayed by shaded G types, but there is no need to stick to this as long as it can be distinguished; for example, it is also possible to distinguish by color.

In a case where the recommended value of a parameter has only one inequality sign, there are two types of distinctions, but the present invention can be implemented by the same processing as in the case of three types.

FIG. 7 is another example of expression of the present invention. This will also be explained using the case of the ratio of kanji included in the readability of a text as an example. 35 is an area indicating a range of values smaller than the recommended value, 36 is an area equal to the recommended value, and 37
is an area larger than the recommended value, 38 is a symbol indicating the actual evaluation value, and 39 is its numerical value.

In the case of this figure, the range of recommended values for the parameters is divided and the evaluation values are also clearly shown, making it easy to judge whether the evaluation text is appropriate or what kind of problems it contains. It's getting easier.

FIG. 8 is an example of another display of the present invention. This is the fifth
FIG. 3A is a diagram in which recommended values of parameters are displayed in parallel with FIG. For comparison, the numbers in FIG. 5(a) are also shown. 40 represents recommended values of parameters.

FIG. 9 is a schematic flow diagram of the present invention. We have proposed various display methods, but here we will introduce the method shown in Figure 6 (a).
) will be explained using the case as an example.

First, the evaluation text 8 stored in the means 2 is called up on the display device 6 of the device. If there is no text, input it using means 1. This evaluation text display processing block is designated as 41.

Next, language analysis processing is performed to generate linguistic information that is a parameter for expressing the readability of the text. This linguistic information creation block is designated as 42.

Next, the evaluation text is analyzed using each of the methods described in 1, 3, 4, and 5, and locations where proofreading is possible are extracted. In this process, a calibration analysis process is executed.

This calibration analysis processing block is designated as 43.

Next, from the processing results of the morphological analysis and the calibration analysis, actual measured values of each parameter representing readability are obtained and stored in a buffer in the control means. This processing block is assumed to be 44.

Next, the evaluation values in the buffer are expressed in the bar graph of FIG. 4, and the evaluation values themselves are displayed. This processing block is assumed to be 45.

Next, the actual measured values and recommended values of the evaluation parameters are compared. This processing block is assumed to be 46. This comparison shows that 3
Branch into 〇 For each branch, change the bar graph display method, for example, A, B, C.

The A and BSC classification display is 47.48.
49.

This method of displaying the classification may be described as a procedure in the control means, or it can be realized by searching for a procedure in a table format using the classification display type ASBSC as a key.

50 is the end processing block - 0 This processing also includes processing to prepare for the evaluation of the next sentence, and is executed after being repeated until the evaluation processing mode is removed or there are no more sentences to evaluate. .

An example of a flow different from the above flow is described below. Above, the sentences are analyzed in the order of 42.43. This was decided because the results of language analysis are often used for proofreading analysis, and any parameter value can be used.

<Effects of the Invention> An advantage of the present invention is that when evaluating a text having recommendable parameter values, not only the evaluation value but also the evaluation value is displayed in the form of a graph, making it easy to see.

It is also effective in that by comparing the evaluation value and the recommended value, it is possible to easily check what kind of qualitative evaluation the evaluation text has received with respect to the parameter.

Based on these evaluation results, it is possible to easily determine which parameter of the text is a problem or the degree of recommendation, and it is also effective in that the text can be modified or added accordingly.

[Brief explanation of the drawing]

Figure 1 is a purchasing block diagram of the device of the present invention, Figure 2 is a diagram showing an example of an evaluation text, Figure 3 is a diagram showing the structure of the buffer and the values of each item, and Figure 4 is a diagram showing the readability of the text. Figures 5(a), 5(b), and 5(c) are diagrams showing examples of the results of evaluating sentences with different proportions of kanji in the readability parameter. Figure 6 (a), (b)
, (c) are diagrams showing examples of displaying the degree of parameters according to evaluation values, FIGS. 7 and 8 are diagrams showing display examples of other embodiments of the present invention, and FIG. FIG. 1 is a schematic flow diagram of the present invention.

Claims (1)

[Claims] A means for inputting/editing Japanese, a means for storing the input Japanese, a means for storing a dictionary, a means for storing grammar, and a method for editing characters and characters to be proofread from the input Japanese. Evaluate a document in a document processing system that has a means for extracting a symbol string, a means for displaying the text and the candidate character/symbol string, and a means for correcting the character/symbol string when there is a character/symbol string to be proofread. 1. A document creation/proofreading support device that distinguishes and displays measured data based on the relative relationship between reference values and measured values of parameters.