JPH11296502A - Method and device for analyzing literature - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a completely new analyzing method and analyzing device for literature which can objectively analyze the sentence style structure of very subjective literature by using computing equipment such as a computer and a spatial theory method which is a mathematical method as a well-worn device and was not seen before. SOLUTION: This is a method for analyzing literature, and three kinds of linear independent element data that the literature has, are subjected to an orthogonal component extracting process (S1) using Gram-Schmidt relation, a normalizing process (S2), and a multi-resolution analyzing process (S3) using wavelet conversion; and three-dimensional data including the obtained three kinds of normalized orthogonal system linear independent element data as (x)-, (y)-, and (z)-directional components are obtained at each resolution level (S4) and represented in a four-dimensional space.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for analyzing a literary work. More specifically, the invention of this application is an analysis of a completely new literary work that can be used to analyze a literary work objectively by a conventional linear space theory method using a computing device such as a computer. The present invention relates to a method and an analyzer.

[0002]

2. Description of the Related Art In general, a literary work is composed of character data, which is a visual representation, as minimum constituent data. The word data obtained by combining character data having a specific meaning is the second constituent data, the word data is combined to form the sentence data as the third constituent data, and the sentence data is combined to form the narrative. Literary works with are formed.

[0003] Conventionally, literary works having such a structure have been evaluated from small structure data to large structure data, such as narratives, composition method of sentence data, and usage of word data. Is performed subjectively from the perspective of This subjective evaluation is not performed numerically, but based on parameters that are difficult to handle numerically, such as sentence data around word data and atmosphere. For this reason, the knowledge of the literary
Different evaluations are made depending on experience, way of thinking, etc.

[0004] However, with regard to objective analysis of literary works, rather than such subjective evaluations, most studies and developments have been carried out, except for the study of using computer to interpret word data in an information science manner. However, highly objective analysis by a computational scientific analysis method using a mathematical method as a usual means has not yet been realized. Therefore, the invention of this application has been made in view of the circumstances described above,
It is possible to objectively analyze the stylistic structure of a literary work with a strong subjective element, using a computer or other computing device, and using a linear spatial theory method, which is a mathematical method, as a conventional means.
It is an object of the present invention to provide an analysis method and an analysis device for a completely new literary work that has never existed before.

[0005]

Means for Solving the Problems The invention of this application is a method for analyzing a literary work, which solves the above-mentioned problems, and comprises a method for analyzing three types of primary independent element data of a literary work. The orthogonal component extraction process using the Gram-Schmidt relationship, the normalization process, and the multi-resolution analysis process using the wavelet transform are sequentially performed, and for each resolution level, the obtained three types of orthonormal primary independent element data are respectively obtained. A literary work analysis method comprising: obtaining three-dimensional data in which x is the x, y, and z-direction components, and expressing the three-dimensional data in four-dimensional space-time (claim 1).
I will provide a.

Further, the present invention provides the above method,
The orthogonal components are extracted from each of the three types of primary independent element data using the Gram-Schmidt relationship, each orthogonal component is normalized to calculate orthonormal primary independent element data, and each orthonormal primary independent element data is waveletted. The multi-resolution analysis using the transform is used to calculate each orthogonal orthogonal primary independent element data for each resolution level, and for each resolution level, each orthogonal orthogonal primary independent element data is calculated as x, y, and z direction components. The present invention provides, as an aspect thereof, calculating the calculated three-dimensional data and converting the three-dimensional data into a four-dimensional spatiotemporal space for each resolution level.

Further, the present invention relates to an apparatus for analyzing a literary work, comprising orthogonal component extraction processing means for extracting an orthogonal component from each of three types of primary independent element data of the literary work using the Gram-Schmidt relationship. , Normalization processing means for normalizing each orthogonal component to calculate orthonormal primary independent element data, and multi-resolution analysis of each orthonormal primary independent element data using wavelet transform to perform normalization for each resolution level. Multi-resolution analysis processing means for calculating orthogonal primary independent element data, and three-dimensional data for calculating three-dimensional data in which each orthonormal primary independent element data has x, y, and z components for each resolution level A literary work analysis device (contractor) comprising calculation means and four-dimensional space-time conversion means for converting three-dimensional data into four-dimensional space-time for each resolution level. Item 3), and an apparatus for analyzing a literary work, using input storage means for storing three types of primary independent element data of the literary work, reading each primary independent element from the storage means, and using a Gram-Schmidt relation from each of the primary independent elements Orthogonal component extraction processing means for extracting orthogonal components, orthogonal component storage means for storing each orthogonal component, reading each orthogonal component from the orthogonal component storage means, and normalizing each of them to calculate orthonormal orthogonal primary independent element data. Normalization processing means, orthonormal data storage means for storing each orthonormal primary independent element data, and reading each orthonormal primary independent element data from the orthonormal data storage means to perform multi-resolution analysis using wavelet transform. A multi-resolution analysis processing means for calculating each orthonormal primary independent element data for each resolution level; Resolution level data storage means for storing crossed primary independent element data, and reading each orthonormal primary independent element data for each resolution level from the resolution level data storage means, and reading each orthonormal primary independent element for each resolution level X,
three-dimensional data calculating means for calculating three-dimensional data as y and z direction components; three-dimensional data storing means for storing three-dimensional data for each resolution level; and three-dimensional data for each resolution level from the three-dimensional data storing means. There is also provided a literary work analysis apparatus (claim 4) comprising: four-dimensional spatio-temporal conversion means for reading original data and converting three-dimensional data into four-dimensional spatio-temporal for each resolution level.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION
As described above, for each of the three types of primary independent element data of the literary work, sequentially, perform orthogonal component extraction processing using Gram-Schmidt relations, normalization processing, and multi-resolution analysis processing using wavelet transform, At each resolution level, three-dimensional data is obtained by using the obtained three types of orthonormal primary independent element data in x, y, and z-direction components, and the three-dimensional data is expressed in four-dimensional space-time. is there.

More specifically, as exemplified in FIG.
First, the input three types of primary independent element data are applied to the Gram-Schmidt relation, and the Gram-Schmidt relation is calculated to extract orthogonal components from each primary independent element data (step 1). Is subjected to normalization processing to calculate orthonormal orthogonal primary independent element data (step 2). Each of the calculated orthogonal orthogonal primary independent element data is subjected to multi-resolution analysis using a wavelet transform to perform calculation. Calculate each orthonormal primary independent element data for each resolution level (step 3), and further use each orthonormal primary independent element data for each resolution level to calculate each orthonormal primary element for each resolution level. Calculate three-dimensional data using the independent element data as components in the x, y, and z directions (step 4). The converted to space when four-dimensional (Step 5), and outputs represented in the spatial time four-dimensional to three-dimensional data of each resolution level. As described above, according to the present invention, the style of a literary work having a strong subjective element can be analyzed very easily and highly objectively by using a computer. Hereinafter, embodiments will be described with reference to the accompanying drawings, and embodiments of the present invention will be described in further detail.

[0010]

Embodiment 1 A literary work analysis method according to the invention of the present application uses a computer to identify a person with modern thinking (hereinafter referred to as a “modern person”) in “Yubisou” by Natsume Soseki. Frequentness, intermediate thinking, that is, frequent occurrence of a person with a thought that does not belong to any of the modern and classical (hereinafter referred to as intermediate type), and a person with classical thinking (hereinafter referred to as a classical person) ) Are selected as three types of primary independent element data related to each other, and the primary individual element data, which are frequently occurring degrees of each person, are objectively analyzed.

The frequency of occurrence of a modern person is as illustrated in FIG. 2 in which the horizontal axis indicates the chapter number and the vertical axis indicates the degree (%). The chapter is a little small, and it gets a little bigger toward the end, but it is convergent. In addition, the frequent degree of the intermediate type person is as illustrated in FIG.
It can be seen from Chapter 1 that there is a tendency toward vibration.

On the other hand, the frequency of occurrence of classical characters is as illustrated in FIG. 4, and the feature is that the change is reduced in the middle chapters 8, 9, and 10. Here, when these FIGS. 2 to 4 are overlapped and collectively expressed, they are as illustrated in FIG. 5. From FIG. 5, it can be seen that there is a certain degree of correlation between the frequency of appearance of each person.

For example, the angle between the frequent degree vector data of a modern person (hereinafter referred to as modern person vector data) and the frequent degree vector data of an intermediate person (hereinafter referred to as intermediate person vector data) is 47.38
It is 37 degrees, and the angle between the frequent appearance degree vector data of the classic type person (hereinafter, referred to as classic type person vector data) and the intermediate type person vector data is 27.3445. Therefore, in the literary work of "Yu-Beauty Grass", the classical character is drawn as being close to the intermediate character.
The angle between modern and classic characters is 53.898
8, which is clearly treated as a conflicting person.

[0014] However, although only a certain degree of correlation can be known from FIG. 5 alone, in more detail, it is objective to see how the person of each thought type appears in the whole "Kyo-Beautiful Grass". Is very difficult to analyze. Therefore, by using the method of the present invention, the frequent appearance of each of the modern-type person, the intermediate-type person, and the classic-type person in “Yu-Beauty-grass” is analyzed in more detail and objectively.

First, as preprocessing, each vector data is transformed into vector data having the number of powers of 2 by adding zero elements. In the first embodiment, the number of elements of the frequent appearance degree vector data of each person is 1
Since there are nine, 13 zero elements are added to the end and 3
The number of two elements is set. An orthogonal component extraction process using a Gram-Schmidt relationship is performed on the frequent appearance vector data of each person transformed into a vector having 32 element numbers in such preprocessing.

More specifically, in the first embodiment, a component orthogonal to the intermediate person vector data is extracted from each of the modern person vector data and the classic person vector data by using the Gram-Schmidt relation. Obtain orthogonal component vector data and second orthogonal component vector data. In order to confirm whether these first orthogonal component vector data and second orthogonal component vector data were correctly extracted, the angle between each vector data and the intermediate type human vector data was calculated. Angle is 9
Since it was 0 degrees, it can be seen that the extraction processing was correctly performed.

Also, since the angle between the first orthogonal component vector data and the second orthogonal component vector data is calculated to be 90 degrees, the first orthogonal component vector data and the second orthogonal component vector data are It can be seen that they are orthogonal to each other. Next, the first orthogonal component vector data, the second orthogonal component vector data, and the intermediate type person vector data are respectively normalized. Each of the obtained vector data is orthonormal modern person vector data,
These are referred to as orthonormal orthogonal classic person vector data and orthonormal intermediate human vector data.

By this normalization processing, the first orthogonal component vector data generated in the orthogonal component extraction processing,
The bias of each of the second orthogonal component vector data and the intermediate type person vector data is removed. Here, three-dimensional vector data in which each of the orthonormal vector data is made into x, y, and z-direction components of a three-dimensional space will be calculated. FIG. 6A illustrates this three-dimensional vector data. In FIG. 6A, the axial direction x is orthonormal intermediate person vector data,
The width direction y is orthonormal orthogonal modern person vector data, and the height direction z is orthonormal classic human vector data. FIGS. 6B, 6C, and 6D show the sizes of the orthonormal intermediate human vector data, the sizes of the orthonormal modern human vector data, and the orthonormal classic human vector data, respectively. This is an example of the size.

As apparent from FIGS. 6A to 6D, the size of the three-dimensional vector data is small in the initial chapter, becomes maximum in the middle chapter, and decreases as approaching the final chapter. You can see that there is. In the middle chapter, not only the vector size is simply increased, but also modern person vector data and classic person vector data appear alternately. In other words, it is possible to objectively know that the person having the modern thinking and the person having the classic thinking alternately appear to constitute the mountain of “Yu-Beauty-grass”.

However, in the three-dimensional vector data illustrated in FIGS. 6A to 6D, there is still a variation occurring in the orthogonal component extraction processing, and this variation is more detailed and accurate objective analysis is performed. Is a factor that hinders Therefore, in the analysis method of the present invention, a multi-resolution analysis process using a wavelet transform is further performed in order to remove the variation and perform a more detailed and highly accurate objective analysis.

Specifically, first, the orthonormal intermediate type human vector data, the orthonormal modern human vector data, and the orthonormal classic human vector data obtained after the normalization processing are each subjected to wavelet transform. Calculate wavelet spectrum data. The obtained wavelet spectrum data is referred to as an intermediate-type person wavelet spectrum and a modern-type person wavelet spectrum, respectively, as a classical-type person wavelet spectrum.

Next, the intermediate-type person wavelet spectrum, the modern-type person wavelet spectrum, and the classic-type person wavelet spectrum are subjected to inverse wavelet transform for each resolution level to perform multi-resolution analysis. The above-described variation can be removed by the multi-resolution analysis processing using the wavelet transform and the inverse wavelet transform.

Then, by using the orthonormal intermediate human vector data, the orthonormal modern human vector data, and the orthonormal classic human vector data for each resolution level, each of the x, y, and z direction components is used. Is created for each resolution level, and each three-dimensional vector data is represented in four-dimensional space-time. More specifically, for example, the orthonormal intermediate person vector data, the orthonormal modern person vector data, and the orthonormal classic person vector data for each resolution level are respectively L1 intermediate data according to each resolution level. Type person vector data, L2 intermediate type person vector data ... Ln intermediate type person vector data, L1 modern type person vector data, L2 modern type person vector data ... Ln modern type person vector data, L1 classic type person vector data , L
(2) Classic-type person vector data... Ln classic-type person vector data. Here, n indicates the number of resolution levels.

For example, when the resolution level is set to six, L1 intermediate type person vector data, L2 intermediate type person vector data... L6 intermediate type person vector data,
L1 modern type person vector data, L2 modern type person vector data ... L6 modern type person vector data, L1 classic type person vector data, L2 classic type person vector data ... L6 classic type person vector data.

Then, for each resolution level, L1.
Ln intermediate type person vector data, 1L1... Ln modern type person vector data, L1... Ln classical type person vector data L1. Vector data is obtained, and each L1... Ln three-dimensional vector data is converted into a four-dimensional space-time with a chapter number as a time axis, for example, the progress of a story in a literary work.

FIGS. 7 (a) to 12 (a) show, respectively,
4 illustrates a four-dimensional spatio-temporal representation of L1 to L6 three-dimensional vector data of resolution levels 1 to 6 obtained when there are six resolution levels in the multi-resolution analysis processing. The time axis in this four-dimensional spatiotemporal representation is a chapter number. For reference, FIG.
(B), (c), (d) to (b), (c), and (d) of FIG. 12 illustrate L1-L6 three-dimensional vector data components in x, y, and z directions.

The four-dimensional space-time of the L1 three-dimensional vector data at the resolution level 1 illustrated in FIGS.
It represents the average appearance frequency of each person in the whole "Yu-Beauty Grass". As is clear from this four-dimensional spatiotemporal expression,
It can be seen that "Yu-Beauty Grass" consists of the appearance of an intermediate character as the whole work. The four-dimensional spatiotemporal space of the L2 three-dimensional vector data of the resolution level 2 illustrated in FIGS. 8A to 8D indicates the frequent appearance of each person in the case where the “Kousou Grass” is divided into two parts. Chapter 3 consists of the appearance of intermediate-type characters, and the last three chapters show a tendency that the frequency of materials of all types of people is decreasing.

Further, the four-dimensional space-time of the L3 three-dimensional vector data of the resolution level 3 illustrated in FIGS. The frequency of frequent appearances of intermediate-type characters has increased almost monotonically, and the frequency of frequent appearances of modern-type and classic-type characters has almost the same tendency, increasing in the middle chapter and decreasing in the last chapter. You can see that there is.

The four-dimensional space-time of the L4 three-dimensional vector data at the resolution level 4 illustrated in FIGS.
Represents the degree of frequent appearance of each person in the case of dividing "Yu-Beauty Grass" into five.The degree of frequent appearance of intermediate persons tends to increase with slight vibration, and modern persons and classic persons appear alternately. It can be seen that the vibration decreases from the beginning to the end.

Further, the four-dimensional spatio-temporal space of the L5 three-dimensional vector data of the resolution level 5 illustrated in FIGS. The frequency of appearance of classical characters is dominant, with the first and second chapters having a low frequency of classical characters and increasing rapidly in the next three and four chapters. After that 5
In Chapters 8 to 8, the average frequent occurrence degree is shown. In Chapters 9 to 16, a large increase and decrease are repeated. In the last three chapters, it is analyzed that the increase and decrease decrease to an average frequent appearance degree.

The four-dimensional spatio-temporal representation of the L6 three-dimensional vector data at the resolution level 6 illustrated in FIGS. It is analyzed, and the frequency of appearance of modern and classic characters is dominant. Modern characters and classic characters have opposite frequency of occurrence, vibrating toward the middle chapter and increasing rapidly to the average frequency with basement iron at the end. Analyzed successfully.

As described above, the variation in the orthogonal component extraction processing is reduced by the multi-resolution analysis using the wavelet transform, that is, the wavelet transform and the inverse wavelet transform.
It is possible to objectively analyze the frequency of frequent occurrences of a person with modern thinking, a person with intermediate thinking, and a person with classical thinking in the literary works of "Yu-Beautiful Grass" in more detail and more easily. .

(Embodiment 2) In this embodiment 2, three types of primary independent element data related to each other, "objective speech,""conversation," and "objective speech," The degree of frequent appearance of each of the intermediate speeches that does not belong to the category is selected, and an objective analysis is performed using a computer.

FIGS. 13 to 15 illustrate the frequent occurrences of the objective speech method, the intermediate speech method, and the conversation method, respectively.
The horizontal axis is the chapter number, and the vertical axis is the frequency (%).
As illustrated in FIG. 13, the frequent appearance of the objective speech method is large at the beginning, slightly smaller at the middle chapter, and oscillating again at the end of the chapter. As illustrated in FIG. 14, the frequency of occurrence of the intermediate speech method has a tendency that the vibration amplitude decreases in the middle chapter, but vibrates with a large amplitude after the middle chapter and converges in the last chapter.

It can be seen that the degree of frequent use of the conversation method has such a tendency that the vibration does not converge at the end chapter, as illustrated in FIG. 13 to 15 are collectively represented as shown in FIG. here,
When the correlation between the frequent occurrence degrees of each speech style is obtained, for example, the angle between the intermediate speech style vector data and the objective speech style vector data is 41.1147, and the angle between the intermediate speech style vector data and the speech style vector data is 54. 7439 can be calculated. Further, the angle between the objective speech method vector data and the conversation method vector data is 35.6356. Therefore, it can be seen from these correlation angles that the objective speech vector data and the conversation vector data are the closest directional vectors.

In the analysis method of the present invention, in order to perform an objective analysis of the frequency of frequent use of each speech in more detail, the first embodiment is used.
In the same manner as described above, orthogonal component extraction processing using Gram-Schmidt, normalization processing, and multi-resolution analysis processing using wavelet transform are sequentially performed, and the obtained three-dimensional data is represented in four-dimensional space-time. First, as preprocessing, 1
The original speech vector data composed of 9 elements is transformed into vector data composed of 32 elements, which is a power of 2, by adding a zero element.

Then, orthogonal component extraction processing using the Gram-Schmidt relation is performed on the vector data of each speech. In the second embodiment, first, a component orthogonal to the intermediate speech vector data is extracted from the objective speech vector data using the Gram-Schmidt relation, and first orthogonal component vector data is obtained. Further, a component orthogonal to the intermediate speech vector data is extracted from the speech vector data, and second orthogonal component vector data is also obtained.

Since the angles between the first orthogonal component vector data and the second orthogonal component vector data and the intermediate speech vector data are each 90 degrees, the extraction process is performed with high accuracy. Can be confirmed. Furthermore, the angle between the first orthogonal component vector data and the second orthogonal component vector data is 90 degrees, indicating that they are orthogonal to each other.

Next, normalization processing is performed on the first orthogonal component vector data, the second orthogonal component vector data, and the intermediate speech vector data which have been determined to be orthogonal as described above, and each of them has The bias generated during the orthogonal component extraction processing is removed. The obtained data are referred to as orthonormal objective speech vector data, orthonormal conversation vector data, and orthonormal intermediate vector data, respectively.

Here, also in the second embodiment, the first embodiment is used.
Similarly to FIG. 17, three-dimensional vector data using these orthonormal vector data as axial components x, width y, and height z components of a three-dimensional space was obtained.
It was as illustrated in (a). FIG. 17 (b)
(C) and (d) respectively show the magnitude of the orthonormal orthogonal intermediate speech vector data which is the axial component x, the magnitude of the orthonormal objective speech vector data which is the width component y, and the height direction z. It illustrates the size of the orthonormal conversational vector data as a component.

As is apparent from FIGS. 17 (a) to 17 (d), intermediate speech is dominant from the first chapter to the middle chapter, and conversation is dominant from the second half to the last chapter. I have. Although the objective speech method is getting larger as it approaches the end of the chapter, it is possible to objectively analyze what is said in the conversation method in the hills of this "Yu-Beautiful grass", but this three-dimensional vector data still has There is variation in each vector data generated during the orthogonal component extraction processing.

Therefore, in the same manner as in the first embodiment,
Perform multi-resolution analysis using wavelet transform. Wavelet transform each of orthonormal objective vector data, orthonormal conversation vector data, and orthonormal intermediate vector data to obtain objective wavelet spectrum, conversation wavelet spectrum, and intermediate speech wavelet spectrum data. After the calculation, each wavelet spectrum is subjected to inverse wavelet transform to perform multi-resolution analysis.

Then, for each resolution level L1... Ln, the obtained L1... Ln orthonormal intermediate person vector data, L1. L1... Ln three-dimensional vector data using the Ln orthonormal system classical person vector data as x, y, and z-direction components is created, and further expressed in four-dimensional space-time.

In the second embodiment, L1... L6 three-dimensional vector data is represented in four-dimensional space-time with six resolution levels. The time axis in this four-dimensional space-time is a chapter number. The L1 three-dimensional vector data at the resolution level 1 exemplified in FIGS. 18B to 18D, that is, the average frequent appearance of each speech style of the “Kyo-Beautiful grass” is shown in FIG.
As shown in (a), it can be expressed in four-dimensional spatio-temporal, and it can be seen from this four-dimensional spatio-temporal expression that the work as a whole is composed of intermediate speech and objective speech.

The frequent appearance of each speech when L2 three-dimensional vector data of resolution level 2 exemplified in FIGS.
The four-dimensional spatiotemporal representation as illustrated in (a) is used. Most of the chapters are composed of the intermediate speech, but in the last three chapters, the frequency of the intermediate speech is reduced.

The L3 three-dimensional vector data of the resolution level 3 exemplified in FIGS. 20B to 20D, that is, the frequency of frequent appearance of each speech when the “Kyoubi Grass” is divided into three parts is shown in FIG.
The four-dimensional spatio-temporal expression as illustrated in FIG. 7A is used, and the objective speech method is becoming dominant instead of the intermediate speech method. Objective speech tends to begin at the beginning, decrease in the middle, and increase again as the end of the chapter approaches.

The L4 three-dimensional vector data of the resolution level 4 exemplified in FIGS. 21B to 21D, that is, the frequency of frequent appearance of each speech when the “Kyo-Beautiful Grass” is divided into five parts is shown in FIG.
The four-dimensional spatiotemporal representation as illustrated in (a) is obtained, and the conversation method becomes dominant, followed by the intermediate speech method. In the conversation method, sudden increase and decrease oscillations occur after the middle chapter, and converge to a constant value at the end chapter.

The L5 three-dimensional vector data of the resolution level 5 exemplified in FIGS. 22B to 22D, that is, the frequent appearance of each speech in each of the two chapters of "Kyoubi Grass" is shown in FIG. The four-dimensional spatio-temporal representation is as shown, and the frequent degree of conversational style is dominant. The frequent degree of conversational style is slightly smaller in the first 1 to 4 chapters, and further decreases in the next 5 to 8 chapters. , The amplitude of the increase / decrease becomes large in the following chapters 9 to 12, and the larger increase / decrease is repeated in the chapters 13 to 16; Become a target. This is 13-16
This means that the chapter forms the hills of this work, and is similar to the tendency of the classic type person in the thought type-based person analysis in Example 1 described above.

The L6 three-dimensional vector data of the last resolution level 6 illustrated in FIGS. 23 (b) to 23 (d), that is, the frequent appearance of each speech style in each chapter of “Kyoubi Grass” is shown in FIG. As shown in (a), it is expressed in four-dimensional spatio-temporal space. From this four-dimensional spatio-temporal expression, the frequent degree of objective speech is dominant, and there are large oscillatory changes after the middle chapter. It can be objectively analyzed that it tends to converge as it approaches.

As a result of objectively analyzing Natsume Soseki's "Yu-Beautiful Grass" by the method of analyzing a literary work of the present invention in the above-described first and second embodiments, the analysis of the characters according to their thought type shows that Classic type characters appear alternately,
The frequency of occurrence increases oscillating towards the middle chapter,
Afterwards, it was analyzed that the degree of average frequent appearance toward the end of the chapter, and in the analysis by speech style, it was possible to analyze that the conversation style constitutes the mountain of the work. It was also found that the conversation style and the classic type person had the same frequent appearance.

As a basis function used for the wavelet transform and the inverse wavelet transform in the analysis method of the present invention, for example, a Daubecies quadratic basis or a Haar basis can be used. In the above-described first and second embodiments, the quadratic basis of Debyssey is used. (Embodiment 3) The analysis of a literary work by the method of the present invention described above can be performed by, for example, an analyzer as illustrated in FIG.

The literary work analysis apparatus of the present invention illustrated in FIG. 24 includes an orthogonal component extraction processing means (1), a normalization processing means (2), a multi-resolution analysis processing means (3), and a three-dimensional data calculation means (4). ) And four-dimensional space conversion means (5). The orthogonal component extraction processing means (1) selects, for example, three types of primary independent element data input from an input device (not shown), for example, the above-mentioned frequent occurrence degree vector data of a person for each thought type. This is applied to the Gram-Schmidt relation, and the Gram-Schmidt relation is calculated to extract the orthogonal component from each linear independent element data.

Each of the orthogonal components received from the orthogonal component extraction processing means (1) is subjected to normalization processing by the normalization processing means (2) to obtain each orthogonal orthogonal primary independent element data. Next, the multi-resolution analysis processing means (3) applies the multi-resolution analysis using the wavelet transform to each orthonormal primary independent element data received from the normalization processing means (2) in the same manner as the processing in the above-described method. After processing is performed, resolution levels L1... L
L1... Ln orthonormal first-order independent element data of each of the three types multiplex-analyzed is calculated.

Further, the three-dimensional data calculating means (4) obtains each of the resolution levels L1... Using the respective orthogonal orthogonal primary independent element data for each of the resolution levels sent from the multi-resolution analysis processing means (3). For each Ln, three types of L
1... Ln three-dimensional data is obtained in which the Xn, y, and z-direction components of each of the Ln orthonormal orthogonal primary independent element data are obtained.

Then, the four-dimensional spatiotemporal conversion means (5) outputs L1... Ln from the three-dimensional data calculation means (4).
The three-dimensional data is represented by being converted into a four-dimensional spatiotemporal space with a story axis such as a chapter number in a literary work as a time axis. The four-dimensional spatio-temporal representation thus obtained by the analysis apparatus of the present invention is displayed on a display device such as a CRT, and is used in the first and second embodiments.
As mentioned above, literary works can be easily and objectively analyzed.

FIG. 25 is a block diagram showing another example of the literary work analyzing apparatus according to the present invention. The literary work analysis device of the present invention illustrated in FIG. 25 includes various storage means for storing each data in addition to each means in the literary work analysis device of FIG. More specifically, three types of primary independent element data input from an input device (not shown) are first stored in the input storage means (11).

The orthogonal component extraction processing means (1) selectively reads out three types of primary independent element data from the input storage means (11) and applies them to the Gram-Schmidt relationship. Then, this Gram-Schmidt relationship is calculated to extract orthogonal components from each primary independent element data. Each of the extracted orthogonal components is sent to and stored in the orthogonal component storage means (12).

The normalization processing means (2) reads out each orthogonal component from the orthogonal component storage means (12), performs normalization processing on each of them, calculates each orthogonal orthogonal primary independent element data, and stores this. Orthogonal data storage means (1
Send to 3). Next, multi-resolution analysis processing means (3)
Performs multi-resolution analysis processing using wavelet transform on each orthonormal primary independent element data read from the orthonormal data storage means (13) in the same manner as the processing in the above-described method to obtain resolution levels L1. Calculate L1... Ln orthonormal primary independent element data of each of the three types multiplexed and analyzed in Ln.

Each of these orthonormal primary independent data for each resolution level is sent to and stored in the resolution level data storage means (14). Further, the three-dimensional data calculating means (4) uses each orthonormal orthogonal primary independent element data for each resolution level read from the resolution level data storage means (14), and uses each of the resolution levels L1.
.. L for each of the three types of L1... Ln orthonormal primary independent element data in x, y, and z direction components
1. Create Ln three-dimensional data.

Each created three-dimensional data is stored in the three-dimensional data storage means (15). Then, the four-dimensional spatio-temporal conversion means (5) converts the L1... Ln three-dimensional data read from the three-dimensional data storage means (15) into a four-dimensional data with a story axis such as a chapter number in a literary work as a time axis. Convert to space-time. The three-dimensional data in the four-dimensional spatiotemporal space obtained in this way is stored in, for example, the output storage means (16).
After being stored in, for example, it is output and expressed.

The various data obtained by the respective means are stored in the respective storage means, so that the data after each processing, calculation and conversion can be displayed and output later, and other devices and the like can be used. Or to compare data from
It is very convenient and makes better use of computer hardware resources. On the other hand, in the analyzer of FIG. 24, the operation of storing data in the storage means and the operation of reading data from the storage means are omitted, so that the analysis processing can be completed at higher speed.

The apparatus of the present invention exemplified in FIGS. 24 and 25 is, for example, a transform in which one or more types of wavelet transform matrices and inverse wavelet transform matrices used for wavelet transform and inverse wavelet transform are stored in advance. Matrix storage means (not shown) may be provided. In this case, the multi-resolution analysis processing means (3) respectively reads out a wavelet transform matrix and an inverse wavelet transform matrix selectively from the transform matrix storage means. To perform wavelet transform and inverse wavelet transform.

Further, a basis function storage means (not shown) in which one or more kinds of basis functions are stored in advance may be provided. In this case, the multi-resolution analysis processing means comprises a basis function storage means. , And a wavelet transform matrix and an inverse wavelet transform matrix are created using the basis functions to perform wavelet transform and inverse wavelet transform.

In selecting these wavelet transform matrices or basis functions, desired ones can be selected by input signals from input means such as a keyboard, a numeric keypad, and a mouse. Of course, the present invention is not limited to the above-described example, and it goes without saying that various aspects are possible in detail.

[0065]

As described in detail above, the method and apparatus for analyzing a literary work of the present invention make it possible to analyze the stylistic structure of a literary work having a strong subjective element very easily and highly objectively. Using a computing device such as a computer, it is possible to develop a completely new field of objective analysis of literary works in accordance with a computational science method that usually uses a mathematical method.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a literary work analysis method of the present invention.

FIG. 2 is a diagram exemplifying the degree of frequent appearance of a modern person.

FIG. 3 is a diagram exemplifying the degree of frequent appearance of an intermediate person.

FIG. 4 is a diagram exemplifying the frequency of appearance of classical characters.

FIG. 5 is a diagram showing an example in which the frequent appearances of a modern person, an intermediate person, and a classic person are superimposed and displayed.

FIGS. 6 (a), (b), (c), and (d) show three-dimensional vector data before multi-resolution analysis processing using wavelet transform, and x,
It is the figure which illustrated the y and z direction component.

7 (a), (b), (c), and (d) are four-dimensional spatial representations of L1 three-dimensional vector data at a resolution level 1, respectively.
FIG. 3 is a diagram illustrating x, y, and z direction components of L1 three-dimensional vector data.

8 (a), (b), (c), and (d) are four-dimensional spatial representations of L2 three-dimensional vector data at resolution level 2, respectively.
FIG. 3 is a diagram illustrating x, y, and z direction components of L2 three-dimensional vector data.

9 (a), (b), (c), and (d) are four-dimensional spatial representations of L3 three-dimensional vector data at resolution level 3, respectively.
FIG. 5 is a diagram illustrating x, y, and z direction components of L3 three-dimensional vector data.

10 (a), (b), (c), and (d) respectively show a four-dimensional space representation of L4 three-dimensional vector data at a resolution level 4 and x, y, and z components of L4 three-dimensional vector data. FIG.

11 (a), (b), (c), and (d) show a four-dimensional space representation of L5 three-dimensional vector data at a resolution level 5 and x, y, and z-direction components of L5 three-dimensional vector data, respectively. FIG.

12 (a), (b), (c), and (d) show a four-dimensional spatial representation of L6 three-dimensional vector data at a resolution level 6 and x, y, and z components of L6 three-dimensional vector data, respectively. FIG.

FIG. 13 is a diagram exemplifying the degree of frequent appearance of the objective speech method.

FIG. 14 is a diagram illustrating an example of a frequent appearance of an intermediate speech method;

FIG. 15 is a diagram exemplifying the degree of frequent use of a conversation method.

FIG. 16 shows an example in which frequent appearance degrees of the objective speech method, the intermediate speech method, and the conversation method are displayed in an overlapping manner.

FIGS. 17 (a), (b), (c) and (d) show three-dimensional vector data before multi-resolution analysis processing using wavelet transform, and x,
It is the figure which illustrated the y and z direction component.

FIGS. 18 (a), (b), (c), and (d) show a four-dimensional spatial representation of L1 three-dimensional vector data at a resolution level 1 and x, y, and z-direction components of L1 three-dimensional vector data, respectively. FIG.

FIGS. 19 (a), (b), (c), and (d) show a four-dimensional space representation of L2 three-dimensional vector data at a resolution level 2 and x, y, and z components of L2 three-dimensional vector data, respectively. FIG.

FIGS. 20 (a), (b), (c), and (d) respectively show a four-dimensional spatial representation of L3 three-dimensional vector data at a resolution level 3 and x, y, and z components of L3 three-dimensional vector data. FIG.

FIGS. 21 (a), (b), (c), and (d) respectively show a four-dimensional space representation of L4 three-dimensional vector data at a resolution level 4 and x, y, and z components of L4 three-dimensional vector data. FIG.

22 (a), (b), (c), and (d) show a four-dimensional spatial representation of L5 three-dimensional vector data at a resolution level 5 and x, y, and z-direction components of L5 three-dimensional vector data, respectively. FIG.

23 (a), (b), (c), and (d) show a four-dimensional spatial representation of L6 three-dimensional vector data at a resolution level 6 and x, y, and z-direction components of L6 three-dimensional vector data, respectively. FIG.

FIG. 24 is a main part configuration diagram showing an example of a literary work analysis device of the present invention.

FIG. 25 is a main part configuration diagram showing another example of the literary work analysis device of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Orthogonal component extraction processing means 2 Normalization processing means 3 Multi-resolution analysis processing means 4 3D data calculation means 5 4D spatiotemporal conversion means 11 Input storage means 12 Orthogonal component storage means 13 Orthogonal data storage means 14 Resolution level data storage Means 15 Three-dimensional data storage means 16 Output storage means

Claims (4)

[Claims] 1. A method for analyzing a literary work, comprising sequentially extracting orthogonal component using Gram-Schmidt relation, normalizing processing, and wavelet for each of three types of primary independent element data included in the literary work. A multi-resolution analysis process using a transform is performed to obtain three-dimensional data using the obtained three types of orthonormal primary independent element data in x, y, and z-direction components for each resolution level. A method of analyzing literary works characterized by representing data in four-dimensional space-time. 2. An orthogonal component is extracted from each of the three types of primary independent element data using the Gram-Schmidt relation, and each orthogonal component is normalized to calculate an orthonormal orthogonal primary independent element data. Each element data is subjected to multi-resolution analysis using wavelet transform to calculate each orthonormal primary independent element data for each resolution level, and for each resolution level, each orthonormal primary independent element data is x, y 2. The method according to claim 1, further comprising: calculating three-dimensional data as components of the z-direction and converting the three-dimensional data into a four-dimensional spatiotemporal space for each resolution level. 3. An apparatus for analyzing a literary work, comprising: orthogonal component extraction processing means for extracting orthogonal components from each of three types of linearly independent element data of the literary work using a Gram-Schmidt relationship; Normalization processing means for normalizing to calculate orthonormal primary independent element data; and multi-resolution analysis of each orthogonal orthogonal primary independent element data using a wavelet transform, and each orthonormal primary independent element for each resolution level. A multi-resolution analysis processing means for calculating data; a three-dimensional data calculation means for calculating three-dimensional data using x-, y-, and z-direction components for each orthonormal primary independent element data for each resolution level; A literary work analysis device comprising: four-dimensional spatio-temporal conversion means for converting three-dimensional data into four-dimensional spatio-temporal for each resolution level. 4. An apparatus for analyzing a literary work, comprising: input storage means for storing three types of primary independent element data of the literary work; reading each primary independent element from the storage means; Orthogonal component extraction processing means for extracting orthogonal components, orthogonal component storage means for storing each orthogonal component, and reading each orthogonal component from the orthogonal component storage means to normalize each to calculate orthonormal orthogonal primary independent element data. Normalization processing means, orthonormal orthogonal data storage means for storing each orthonormal primary independent element data, and reading each orthonormal primary independent element data from the orthonormal data storage means to perform multi-resolution analysis using wavelet transform. Multi-resolution analysis processing means for calculating each orthonormal primary independent element data for each resolution level; Resolution level data storage means for storing orthogonal system primary independent element data; reading out each orthonormal orthogonal primary independent element data for each resolution level from the resolution level data storage means, and reading each orthonormal primary independent element for each resolution level Three-dimensional data calculating means for calculating three-dimensional data using the data as x, y, and z-direction components; three-dimensional data storing means for storing three-dimensional data for each resolution level; A literary work analyzing apparatus comprising: four-dimensional spatio-temporal conversion means for reading three-dimensional data for each resolution level and converting the three-dimensional data into four-dimensional spatio-temporal for each resolution level.