RU2538304C1 - Method for automatic semantic classification of natural language texts - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: method for automatic semantic classification of natural language texts comprises presenting each text to be classified in digital form for subsequent processing; indexing the text to obtain elementary units of the first through fifth levels; detecting the frequency of occurrence of units of the fourth level, each being a semantically significant object or attribute, and the frequency of occurrence of semantically significant relationships linking semantically significant objects, as well as objects and attributes; forming a semantic network from a triad which is units of the fifth level; renormalising the frequencies of occurrence into the semantic weight of the units of the fourth level; ranking the units of the fourth level according to the semantic weight by comparison thereof with a threshold value and those having a weight below the threshold value; detecting the degree of crossing semantic networks of the text and text samples; selecting as a class for text object regions, the degree of crossing the semantic network with the semantic network of text is greater than the threshold.

EFFECT: faster process of comparing texts.

6 cl, 2 dwg, 24 tbl

Description

FIELD OF THE INVENTION

The present invention relates to the field of information technology, and in particular to a method for automated semantic classification of texts in natural language.

State of the art

There are various ways of automated semantic (i.e. semantic) classification of texts in natural languages (see, for example, patents of the Russian Federation No. 2107943 (publ. 03/27/1998) and No. 2108622 (publ. 10.04.1998), as well as the application EPO No. 0241717 (publ. 21.10.1987)).

Generally speaking, the semantic classification of texts in natural language cannot be carried out directly, since in this case it is necessary to classify not by the presence of specific words in the text, but by the meaning behind whole sentences and even paragraphs or sections. Therefore, usually the semantic classification of texts is preceded by the semantic indexing of these texts, which is carried out in various ways. Moreover, the elimination of the semantic ambiguity of these texts is important.

Such methods of semantic indexing of texts for their subsequent comparison with the elimination of semantic ambiguity are described, for example, in RF patent No. 2242048 (publ. 10.12.2004), in US patents No. 6871199 (publ. 22.03.2005), 7024407 (publ. 04.04.2006 ) and 7383169 (publ. 06/03/2008), in applications for US patent No. 2007/0005343 and 2007/0005344 (both publ. 04.01.2007), 2008/0097951 (publ. 04.24.2008), in Japanese applications laid out No. 05 -128149 (publ. 05/25/1993), 06-195374 (publ. 07/15/1994), 10-171806 (publ. 06/26/1998) and 2005-182438 (publ. 07/07/2005), in EPO application No. 858586 (publ. . 07.15.1998).

Closest to the claimed invention can be considered a method of automated semantic indexing of text in a natural language, disclosed in the patent of the Russian Federation No. 2399959 (publ. 09/20/2010). In this method, text in digital form is segmented into elementary units of the first level (words); form for each elementary unit of the first level (words) an elementary unit of the second level (normalized word form); segment text in digital form into sentences corresponding to sections of indexed text; reveal in the text, in the process of linguistic analysis, elementary units of the third level (stable phrases); in the process of multi-stage semantic-syntactic analysis by referring to linguistic and heuristic rules preformed in the database in a predefined linguistic environment, elementary units of the fourth level (semantically significant object and its attribute) and semantically significant relationships between the identified semantically are revealed in each of the sentences significant objects, as well as between semantically significant objects and attributes; form within this text for each of the identified semantically significant relationships, a set of elementary units of the fifth level (triads); index on a set of formed triads all semantically significant objects connected by semantically significant relations, as well as attributes individually and all triads of the form “semantically significant object - semantically significant relation - semantically significant object”, as well as all triads of the form “semantically significant object - semantically significant” meaningful relation - attribute ”; save the generated triads and the resulting indexes in a database together with a link to the source text from which these triads are formed.

The disadvantage of this method is the lack of ranking of the formed elementary units of the fourth level in terms of their relevance to the text, which leads to an unreasonably large amount of calculations associated with the need to use the entire generated index for further processing.

Disclosure of invention

The purpose of the present invention is to expand the arsenal of methods for the semantic classification of texts in natural languages by accelerating the process of comparing texts.

The achievement of this goal and the receipt of the specified technical result is provided in the present invention by means of an automated semantic classification of texts in natural language, which consists in the following: represent each classified text in digital form for subsequent automatic and (or) automated processing; indexing each classified text in digital form, obtaining: elementary units of the first level, including at least words, elementary units of the second level, each of which is a normalized word form, elementary units of the third level, each of which is a stable phrase in text, elementary units of the fourth level, each of which is a semantically significant object and attribute, and elementary units of the fifth level, each ryh triad represents either of the two named entities and named relations therebetween, or from semantically meaningful object and the attribute and linking them semantically meaningful relations; identify the frequency of occurrence of elementary units of the fourth level and the frequency of occurrence of semantically significant relationships; save in the database the formed elementary units of the second, third, fourth and fifth levels with the detected frequencies of occurrence of elementary units of the fourth level and semantically significant relations, as well as the resulting indices, together with links to specific sentences of this text; form a semantic network of triads in such a way that the first elementary unit of the fourth level of the subsequent triad is associated with the same second elementary unit of the fourth level of the previous triad; carry out, in the course of an iterative procedure, renormalization of the frequencies of occurrence into the semantic weight of elementary units of the fourth level, which are the vertices of the semantic network, so that elementary units of the fourth level, connected in a network with a large number of other elementary units of the fourth level with a high frequency of occurrence, increase their semantic weight, and other elementary units of the fourth level evenly lose it; rank elementary units of the fourth level by semantic weight by comparing the semantic weight of each of them with a predetermined threshold value and delete elementary units of the fourth level having semantic weight below the threshold value; retain in memory the remaining elementary units of the fourth level with a semantic weight above the threshold, as well as semantically significant relations between the remaining elementary units of the fourth level; reveal the degree of intersection of the semantic network of classified text and semantic networks of text samples, which text samples are composed of previously classified texts and describe the subject areas of semantic classification, while the degree of intersection is revealed both by the vertices of the semantic networks and by the relationships between these vertices, taking into account semantic weights vertices of the considered semantic networks and weight characteristics of their connections, and take the identified degree of intersection of semantic networks erentiable text and a particular text sample as the quantity characterizing the semantic similarity of the classified text and this text sample; at least one of the subject areas is selected as a class for the classified text, the degrees of intersection of the semantic network of which with the semantic network of the classified text are greater than a predetermined threshold.

A feature of the method of the present invention is that when a predetermined threshold is exceeded by degrees of intersection for several subject areas, subject areas can be ranked by their degree of proximity to the classified text.

Moreover, they can choose a predetermined number of subject areas to which the classified text belongs.

Another feature of the method of the present invention is that indexing is carried out in the process of performing the following steps: segment the text in digital form into elementary units of the first level, including at least words; Segment digitally into sentences according to graphematic rules; form for each elementary unit of the first level, which is a word, based on morphological analysis, elementary units of the second level, including a normalized word form; calculate the frequency of occurrence of each elementary unit of the first level for two or more adjacent units of the first level in this text and combine among the elementary units of the first level the sequence of words following each other in this text into elementary units of the third level, which are stable combinations of words, if, for every two or more consecutive words in a given text, the difference in the calculated frequencies of occurrence of these words for the first occurrence of a given sequence with s and a number of subsequent occurrences for each pair of sequences of words remain unchanged; identify, in the process of multi-stage semantic-syntactic analysis by referring to linguistic and heuristic rules pre-formed in the database in a predefined linguistic environment, in each of the generated sentences semantically significant objects and attributes are elementary units of the fourth level; for each elementary unit of the fourth level, the identity is fixed by reference between the corresponding semantically significant object, as well as the attribute, and the corresponding anaphoric reference, if any, in the classified text, replacing each anaphoric reference with the corresponding antecedent; store in memory every semantically significant object and attribute; reveal, in the process of multi-stage semantic-syntactic analysis by referring to linguistic and heuristic rules pre-formed in the database in a predefined linguistic environment, in each of the generated sentences, semantically significant relationships between identified fourth-level units - semantically significant objects, as well as between semantically significant objects and attributes; assign to each semantically significant relation the corresponding type from the subject ontology stored in the database on the subject of the subject area to which the classified text belongs; identify throughout the text the frequency of occurrence of elementary units of the fourth level and the frequency of occurrence of the mentioned semantically significant relationships; store in memory each identified semantically significant relation together with the type assigned to it; form within the given text for each of the identified semantically significant relations, linking both the corresponding semantically significant objects and the semantically significant object and its attribute, a multitude of triads that are elementary units of the fifth level; on a set of formed triads individually index all semantically significant relations related semantically significant objects with their frequencies of occurrence, all attributes with their frequencies of occurrence, and all formed triads.

Another feature of the method of the present invention is that the degree of intersection of two semantic networks is calculated as the sum of the coincidences of the elementary units of the fifth level of these two semantic networks.

At the same time, the stages are carried out in which: one of the two semantic networks is selected as the basic network, in which, after ranking and removing vertices with semantic weights, more vertices remain below the threshold value than in the other, which is chosen as the comparison one; find for each vertex of the core network in the compared network a vertex that is the same elementary unit of the fourth level, i.e. the same semantically significant object, or the same attribute; calculate, for each vertex found in each of the base and compared networks, the values of all triads associated with a given vertex as the area of triangles whose sides correspond to the components of each of these triads, and the angle between the sides is proportional to the weight of the semantically significant relation; choose for each pair of triads associated with a pair of specific vertices in the base and the compared networks, the smaller of the calculated values as the degree of intersection of the mentioned triads in the base and the compared networks; summarize for each vertex associated with a given vertex all selected calculated values, obtaining the degree of intersection for a given pair of vertices of the base and compared networks; normalize the found amount to the number of semantically significant objects and attributes associated with a given vertex in that of the base and compared networks, which contains more vertices associated with this vertex; summarize the normalized sums for all the vertices of that of the base and compared networks, which contains more vertices; normalize the amount received to the number of elementary units of the fourth level remaining in this network, obtaining the degree of intersection of two semantic networks.

Brief Description of the Drawings

The present invention is further explained by the description of a specific example of its implementation and the accompanying drawings.

Figure 1 shows the conditional block diagram explaining the claimed method.

2 is a flowchart illustrating a preferred method for indexing text.

DETAILED DESCRIPTION OF THE INVENTION

The method of the present invention can be implemented in almost any computing environment, for example, on a personal computer connected to external databases. The steps of the method are illustrated in FIG.

All further explanations are given as applied to the Russian language, which is one of the most highly inflected languages, although the proposed method is applicable to the semantic classification of texts in any natural languages.

First of all, each of the texts subject to semantic classification must be submitted in electronic form for subsequent automated processing. This step in figure 1 is conventionally indicated by the reference number 1 and can be performed in any known manner, for example, by scanning the text and subsequent recognition using well-known means such as ABBYY FineReader. If the text is received for classification from an electronic network, for example, from the Internet, then the stage of its submission in electronic form is carried out in advance, before this text is posted on the network.

Professionals should be clear that the operations of this and subsequent steps are carried out with storing intermediate results, for example, in random access memory (RAM).

The text converted into electronic form is sent for processing, during which indexation is carried out. This indexing (step 2 in FIG. 1) can be done in the same way as it is disclosed, for example, in the aforementioned patent of the Russian Federation No. 2399959 or in the patent application US 2007/0073533 (publ. March 29, 2007). In the process of this indexing, elementary text units of different levels are obtained. Elementary units of the first level include at least words; each of the elementary units of the second level is a normalized word form; each of the elementary units of the third level is a sequence of successive words in the processed text; each of the fourth-level elementary units is a semantically significant object or attribute; each of the elementary units of the fifth level is a triad of either two semantically significant objects and a semantically significant relationship between them, or a semantically significant object and its attribute, and a semantically significant relation connecting them.

It is preferable, however, to index the text using the method claimed in the patent application of the Russian Federation 2012150734 (priority from 11/27/2012) and illustrated in FIG. 2. In this method, the text in digital form is first segmented into elementary units of the first level, including at least words. In the mentioned patent of the Russian Federation No. 2399959, these elementary units of the first level are called tokens. A token can be any text object from the following set: words consisting of each sequence of letters and, possibly, hyphens; sequence of spaces; punctuation marks; numbers. Sometimes sequences of symbols such as А300, i150b, etc. are also included here. Tokens are always allocated according to fairly simple rules, for example, as in the aforementioned RF patent No. 2399959. In Fig.2, this step is conventionally indicated by reference numeral 21.

Following this, at step 22 (FIG. 2), indexed text is digitally segmented into sentences corresponding to portions of the given text. Such segmentation is carried out according to graphematical rules. For example, the simplest rule for highlighting sentences is: “A sentence is a sequence of tokens, starting with a capital letter and ending with a period”.

Further, for each elementary unit of the first level (for each token), which is a word, on the basis of morphological analysis, the corresponding elementary unit of the second level is formed, which is a normalized word form, hereinafter referred to as the lemma. For example, for the word “go” the normalized word form will be “go”, for the word “beautiful” the normalized word form will be “beautiful”, and for the word “wall” the normalized word form will be “wall”. In addition, for each word form, the part of speech to which the given word belongs and its morphological characteristics are indicated. Naturally, for different parts of speech, these characteristics are different. For example, for nouns and adjectives it is a gender (masculine - feminine - average), number (singular - plural), case; for verbs it is a form (perfect - imperfect), person, number (singular - plural); etc. Thus, for a given word, its normalized word form (lemma) + morphological characteristics, including part of speech, are its morph. One and the same word can have several morphs. For example, the word "glass" has two morphs - one for a noun of the middle gender and one for a past tense verb. This step is conventionally indicated in figure 2 by reference numeral 23.

The next step, conventionally indicated in FIG. 2 by reference numeral 24, is that for each of the mentioned first level elementary units, the frequency of occurrence is counted in said text. In other words, they determine how many times each word occurs in the processed text. This operation is carried out automatically, for example, by simply calculating the frequency of occurrence of each token, either as described in the patent of the Russian Federation No. 2164450 (publ. 05/20/2001), or in US patent No. 6189002 (publ. 13.02.2001). Simultaneously with the calculation of the frequency of occurrence, for every two or more words that follow in a given text, the differences in the calculated frequencies of occurrence of these words are found in the first occurrence of this sequence of words and in their subsequent occurrences. If these differences for the first occurrence of a given sequence of words and for several subsequent occurrences of them remain unchanged, such a sequence of words following each other in this text (i.e., elementary units of the second level) are combined into elementary units of the third level, which are stable collocations.

Further, in the next step, indicated in FIG. 2 by the reference numeral 25, in order to identify semantically significant objects and attributes, a multi-stage semantic-syntactic analysis is performed. Such multistage semantic-syntactic analysis is performed by referring to the linguistic and heuristic rules generated in the database in a predetermined linguistic environment. Such a medium may be, for example, the linguistic medium mentioned in the aforementioned application for US patent No. 2007/0073533 or in the above patents of the Russian Federation No. 2242048 and the Russian Federation No. 2399959, or any other linguistic medium that defines the appropriate rules that allow to eliminate syntactic and semantic ambiguities words and expressions of real text. Linguistic and heuristic rules in the selected environment are referred to below as rules.

The identification of semantically significant objects and attributes, which are considered elementary units of the fourth level, is performed in the sentence on the set of elementary units of the first, second and (or) third levels.

For each semantically significant object or attribute, i.e. elementary units of the fourth level with the types assigned to them find the corresponding anaphoric link (if any). For example, in the sentence “Mechanics is a part of physics that studies the laws of mechanical motion and the causes that cause or change this movement”, the anaphoric reference to the word “mechanics” will be the pronoun “which”, while the word “mechanics” will be an antecedent for this anaphora, and also, the anaphoric reference to the word “mechanical” will be the pronoun “this,” while the word “mechanical” will be an antecedent for this anaphora. This step of finding the anaphoric link is conventionally indicated in FIG. 2 by reference numeral 26. Each anaphoric link is replaced with its corresponding antecedent. After that, each identified semantically significant object and attribute is stored in the corresponding memory.

In the next step, indicated by a reference numeral 27 in FIG. 2, a multi-stage semantic-syntactic analysis is performed, with which, on the basis of elementary units of the first, second, third and fourth levels, semantically significant relations between semantically significant objects are found using the above rules, and between semantically significant objects and attributes.

At the step indicated by reference numeral 28 in FIG. 2, each semantically significant relation is assigned a corresponding type from the subject ontology stored in the database on the subject of the subject area to which the indexed text belongs. After that, each semantically significant relation is stored in the corresponding memory together with the type assigned to it and the morphological and semantic attributes found for it.

After that, at the stage indicated by reference numeral 29 in FIG. 2, the occurrence frequencies of semantically significant objects and attributes, as well as the occurrence frequencies of semantically significant relationships between semantically significant objects and between semantically significant objects and attributes throughout the entire text, are detected. This operation is performed in almost the same way as in step 24 for elementary units of the first level.

In the step indicated in FIG. 2 by reference numeral 30, stored semantically meaningful objects, as well as attributes, and semantically meaningful relationships are used to form triads. Moreover, within the limits of the text being indexed, for each of the identified semantically meaningful relationships that connect certain semantically meaningful objects and attributes, many triads of two types are formed. Each of the many triads of the first type includes a semantically meaningful relationship and two semantically meaningful objects that are linked by this semantically meaningful relationship. Each of the many triads of the second type includes a semantically significant relation, one semantically significant object, as well as its attribute, which are associated with this semantically significant relation. If two semantically significant objects are denoted by O _i and O _j , and the semantically significant relation connecting them is denoted by R _ij , then each of the triads of the first type can be conditionally represented (depicted) as O _i → R _ij → O _j . Each of the triads of the second type can be represented as O _i → R _im → A _m , where A _m are the corresponding attribute, and R _im associates the semantically significant object and the attribute is a semantically significant relation. In these entries, the indices i, j, m are integers.

Then, in the step indicated by reference numeral 31 in FIG. 2, the text is indexed. In this case, individually associated with the semantically significant relations, all semantically significant objects with their frequency of occurrence, all attributes with their frequency of occurrence, and all formed triads are indexed individually on the set of formed triads.

To do this, on the set of formed triads, all semantically significant objects and their attributes are separately indexed, with their frequencies of occurrence, and all triads of the form “semantically significant object - semantically significant relation - semantically significant object”, as well as all triads of the form “semantically significant object - semantically significant relation is an attribute. ” The indices generated in step 30 of the triad and the indices obtained in step 31, together with a link to specific sentences of the source text from which these triads are generated, are stored in the database (step 32 in FIG. 2).

For specialists, it is obvious that the storage devices mentioned at separate stages can in fact be both different devices, and one storage device of sufficient volume. Similarly, the individual databases mentioned at the respective stages can be not only physically separate databases, but also the only database. Moreover, said storage devices (memories) may store the same single database, or store said databases separately. Those skilled in the art will also understand that the methods claimed in the present invention are executed in an appropriate computing environment under the control of appropriate programs that are recorded on computer-readable media intended for direct participation in a computer.

Returning to the block diagram of FIG. 1. At stage 3, the frequencies of occurrence of elementary units of the fourth level (i.e., semantically significant objects and attributes) are identified, as well as the frequencies of occurrence of semantically significant relationships are identified. Note that the formed elementary units of the fourth level are stored in the database along with the identified frequency of occurrence. In addition, the resulting indexes are stored in the database along with links to specific sentences of the text.

Then, in step 4, a semantic network is formed in the method of the present invention such that the first semantically significant object of the subsequent triad is associated with the same second semantically significant object of the previous triad. At the same time, in the course of the iterative procedure, the frequency of occurrence of semantically significant objects and attributes is renormalized into the semantic weight of semantically significant objects and attributes, which are the vertices of the semantic network. This renormalization is carried out in such a way that semantically significant objects and attributes connected in a network with a large number of semantically significant objects and attributes with a high frequency of occurrence increase their semantic weight, while other semantically significant objects and attributes evenly lose it (step 5 in FIG. one).

Next, the elementary units of the fourth level are ranked by semantic weight by comparing their semantic weight with a predetermined threshold value (step 6 in FIG. 1).

The elementary units of the fourth level with a semantic weight below the threshold are removed (step 7 in FIG. 1). The remaining elementary units of the fourth level with a weight above the threshold are stored in memory (step 8). The semantically significant relations between semantically significant objects and also between semantically significant objects and attributes remaining in the semantic network are also stored in memory.

Next, at step 9, the degree of intersection of the constructed semantic network of classified text and semantic networks of text samples is revealed. These text samples are composed of previously classified texts. They describe the subject areas of the semantic classification for which the classified text is processed. Moreover, the degree of intersection of semantic networks is revealed both by their vertices and by the relationships between these vertices, taking into account the semantic weights of the vertices of the considered semantic networks and the weight characteristics of their relationships.

The revealed degree of intersection of the semantic networks of the classified text and the specific text selection is taken as a value characterizing the semantic similarity of the classified text and this text selection. After that, at least one of the subject areas, the degree of intersection of the semantic network of which with the semantic network of the classified text are greater than a predetermined threshold, is selected as a class for the classified text (step 10 in FIG. 1).

The degree of intersection of two semantic networks formed in the manner described above is calculated as the sum of the coincidences of the elementary units of the fifth level of these two semantic networks. In principle, this calculation can be carried out by various methods known to those skilled in the art.

и $${\overline{c}}_j$$

, где вектор $${\overline{c}}_i$$

соответствует первому семантически значимому объекту или атрибуту элементарной единицы пятого уровня, вектор $${\overline{c}}_j$$

соответствует второму семантически значимому объекту, либо атрибуту элементарной единицы пятого уровня, а угол между векторами c_i и c_j, равный w_ij, пропорционален частоте встречаемости семантически значимого отношения между первым и вторым семантически значимыми объектами или между первым семантически значимым объектом и атрибутом, нормированной на 90°: w_ij∈(0…90°).Preferably, the degree of intersection can be calculated as the sum of the intersections of the fifth level elementary units of these two networks. For this, one of the two semantic networks is selected as the basic network, in which, after ranking and deleting vertices with semantic weights below the threshold value (see step 7 in FIG. 1), more vertices remain than in the other, which is chosen as the comparison. For each vertex of the core network, a vertex is found in the compared network, which is the same elementary unit of the fourth level, i.e. the same semantically significant object, or the same attribute. For each vertex found in each of the base and compared networks, the values of all triads connected with this vertex are calculated as the area of triangles whose sides correspond to the components of each of these triads. This area calculation can be performed as a 100% normalized scalar product on vectors $${\overline{c}}_i$$

and $${\overline{c}}_j$$

where is the vector $${\overline{c}}_i$$

corresponds to the first semantically significant object or attribute of an elementary unit of the fifth level, the vector $${\overline{c}}_j$$

corresponds to the second semantically significant object, or the attribute of an elementary unit of the fifth level, and the angle between the vectors c _i and c _j equal to w _ij is proportional to the frequency of occurrence of the semantically significant relationship between the first and second semantically significant objects or between the first semantically significant object and the attribute normalized 90 °: w _ij ∈ (0 ... 90 °).

Next, choose for each pair of triads associated with a pair of specific vertices in the base and the compared networks, the smaller of the calculated values as the degree of intersection of the triads in the base and the compared networks. All selected calculated values are summed for each of the vertices, obtaining the degree of intersection for a given pair of vertices of the base and compared networks. The found sum is normalized to the number of semantically significant objects and attributes associated with a given vertex in that of the base and compared networks, which contains more vertices. The obtained normalized sums are now summed over all the vertices of that of the base and compared networks, which contains more vertices. Finally, the resulting total amount is normalized to the number of elementary units of the fourth level remaining in this network, i.e. semantically significant objects and attributes receiving the degree of intersection of semantic networks.

Obviously, if there is no vertex in the compared network, the degree of intersection for this vertex is taken equal to zero.

Example

To illustrate the implementation of the claimed method of automated semantic classification of text in natural language, consider the following example. Let there be some Russian-language text presented on the Internet site http://www.unn.ru/eng/priem.htm, and several (for example, three) samples of texts characterizing the classes (subject areas) presented on the same site. Thus, we can assume that the conversion of texts into electronic form, indicated in FIG. 1 by reference numeral 1, has already been completed.

A typical example of such text is the following snippet:

“Throughout the world, a math exam is a written solution to problems. The written nature of the trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates. Indeed, in an oral exam, a student is completely defenseless. I happened to hear, while taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics of Moscow State University, examiners who drowned students at the next table who gave impeccable answers (possibly exceeding the level of understanding of the teacher). There are also known cases when they drowned on purpose (sometimes it can be saved from this by entering the audience on time). ”

In accordance with the claimed method of automated semantic classification of texts in natural language, a previously created base of syntactic rules and dictionaries is used, within which text processing and construction of a semantic index will be carried out. Such databases are prepared by linguistic experts, who, based on their experience and knowledge, determine the sequence and composition of the syntactic processing of the text specific to a particular language.

Linguistic experts preliminarily construct many syntactic rules that allow using the corresponding linguistic dictionaries also previously constructed by expert linguists to automatically identify specific information in processed texts corresponding to semantically significant objects, attributes of semantically significant objects, and semantically significant relationships that may have space between semantically significant objects or between semantically significant objects Tami and attributes.

In addition to the specification of the subject area and the rules in accordance with the above methods, dictionaries of general and special vocabulary are used.

In accordance with the claimed method of automated semantic comparison of texts in natural language, the text is first segmented into elementary units — tokens (reference position 21 in FIG. 2) and morphological analysis of word tokens (reference position 23 in FIG. 2). As a result of this step, the source text is transformed into many tokens and morphs, which are presented in Table 1 and Table 2, respectively.

Introductory words and plug-in constructions do not carry any syntactical load, therefore tokens of this type are excluded from further analysis.

Geographic tokens are considered as one word, with a morph corresponding to the morph of the main word.

Further, after segmenting the text into tokens and morphological analysis of the word tokens, stable phrases are extracted (reference position 24 in FIG. 2). To do this, calculate the frequency of occurrence of words in sequences of two or more words in the text. Then, the differences in the frequencies of occurrence of words in the sequence are compared for the first appearance of a given sequence of words and for several subsequent occurrences of them.

The frequency of occurrence of words at the first appearance of the sequence and at its subsequent appearance, as well as the difference of these frequencies are presented in Table 3.

As a result of this stage, the source text, in addition to elementary units of the first and second levels, is supplemented by many units of the third level - stable phrases. The phrases for our example are presented in Table 4.

After the above steps are completed, the processed text is fragmented into sentences (reference position 22 in FIG. 2). As a result of this step, the sets formed above are supplemented by the set of proposals presented in Table 5.

Thus, after all the above steps are completed, the processed text will be segmented into sentences, each of which is marked with sets of annotations of elementary units of the first, second and third levels.

Following this, in accordance with the claimed method of automated semantic comparison of texts in natural language, the semantically significant objects and attributes (elementary units of the fourth level) are identified (reference position 25 in FIG. 2). It is made in each sentence on the set of elementary units of the first, second and (or) third levels by applying a pre-formed set of linguistic and heuristic rules using the pre-formed corresponding linguistic dictionaries.

Semantic-syntactic processing of sentences is carried out in several stages. We will carry out all stages on the text that we have chosen as an example.

1. Subdivision of sentences on punctuation marks and unions (union words and phrases) into initial fragments and determining the type of fragment based on its morphological characteristics. For this, a dictionary of unions, union words and phrases is used.

Borders of fragments are set for all punctuation marks and unions (union words and phrases) without a comma. In addition, the dictionary of unions determines whether there is such a complex union, the beginning of which is in the fragment to the left of the left, and the end is in this. In our case, such an allied phrase is “as long as”. If there is such a union, then a comma is transferred before the whole union.

The fragment type is one of the following values indicated in table 6. In the order indicated in table 6, the word form with the corresponding homonym is searched in the fragment, the remaining homonyms of the found word form are not considered.

2. Combining the original segments with simple cases of homogeneous series of adjectives, adverbs, nouns, etc. A sign of homogeneity is the presence of a compositional union (or comma), before and after which there should be word forms of one part of speech, which have homonyms that have the same morphological information. The remaining homonyms are not considered during further analysis, thus, partial removal of homonymy occurs.

In our example, segments 2.1 and 2.2 are connected by a “like” conjunctive union, since the tokens 14 (“character”) and 26 (“choices”) of Table 1 have homonyms for one part of speech that have the same morphological information - Im.p. or Win.p. The type of segment received is 1.

3. Construction of simple syntactic groups corresponding to the attribute level of description (Table 8): attribute of an object / subject / action + object / subject / action, measure of an attribute of an object / subject / action + object / subject / action.

Further in sentences of the text anaphoric links are revealed and revealed. To do this, within the entire text being processed, during the stage indicated in FIG. 2 by the reference numeral 26, pronouns are found that can be anaphoric references to the corresponding words, and for pronouns that really are, they fix the identity by reference between the corresponding semantically significant object and its anaphoric reference. In our example, there are no anaphora.

4. Attachment of contact-located fragments (participial, participial revolutions, accessory definitive, etc.) and establishing a hierarchy on fragments. The participial turnover and the adjunctive definitive will be a sign of the corresponding object, the participial turnover - a sign of action.

In our example, the following attachments are made:

- a fragment 4.2 (Table 7) with type 6 “taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics” is a participle with the main word “accepting”, therefore, the whole fragment 4.2 obeys the verb of the previous fragment “hear”,

- fragment 4.5 (Table 8) with type 5 “giving perfect answers” is the sacrament with the main word “giving”, consistent with the noun “students” of the previous fragment by gender and number, therefore, the whole fragment 4.5 is subordinate to the noun “students”, being its indicative description. Thus, the whole fragment 4.5 is an attribute (attribute) of the noun “students”.

The second column of table 10 shows the enlarged fragments of the proposal received after the attachment.

5. The construction of many unique morphological interpretations of each fragment.

Within each sentence, partial removal of homonymy is carried out at the morphological level by:

1) the allocation of groups of nouns consistent with one or more adjectives / participles / pronouns-adjectives that are in homogeneous connection (the so-called attributive level described above in clause 3);

2) analysis of the location of the dash, which removes homonymy, firstly, from the word of the form “this”, since the dash before this wordform indicates that “this” is a particle, and secondly, from nouns before and after the dash, etc. to. the closest to the dash of a noun on the right is only a nominative case, and on the left is a nominative or instrumental. So, in our example, the word forms “this” (token 8, Table 2) are particles, and the word forms “exam” (token 4, Table 2) and “solution” (token 10, Table 2) can only be used in accusative case;

3) the identification of participial phrases following the noun and participle phrases, since such phrases are separated by commas, and nouns included in them depend on the verb form and cannot be in the nominative case. So, in our example, the word forms “exam” (token 45, Table 2) and “answers” (token 65, Table 2) cannot be in the nominative case;

4) identifying prepositions, while those homonyms that have a case not used with this preposition are removed from the subordinate of the preposition of the noun (the preposition management model is used). In our example:

- the preposition “from” (token 27, Table 1) before the word form of “candidates” (token 29, Table 1) cannot manage a noun in the accusative case;

- the preposition “on” (token 46, table 1) before the word form “department” (token 47, table 1) cannot control a noun in the dative case;

- the word form "me" (token 40, Table 1), before which there is no excuse, cannot have a prepositional case,

therefore, these homonyms are removed from consideration.

In table 2, variants of homonyms that are excluded from consideration as a result of partial removal of homonymy at the morphological level are highlighted in gray.

6. Combining fragments into simple sentences as part of a complex subject using subordinate unions. Subordinate unions act as boundaries of simple sentences (Table 10, column 3).

7. Identification of the predicative minimum (including the main semantically significant objects, and the main semantically significant relationships - predicates) of a sentence by comparing its structure with a dictionary of templates of minimal structural sentence schemes, a fragment of which is shown in Table 11. The result for our example is given in Table 12.

8. The selection of the remaining members of a simple sentence (other semantically significant objects and attributes) and other semantically meaningful relationships is carried out by a consistent comparison of the words of the sentence with the actual structure of the verb from the dictionary of valencies of verbs. The filled valence nests for predicates of the example text are shown in Table 13.

We consider the predicate stoked in more detail. According to the semantic classification used in the dictionary of valencies of verbs, he predicts the situation of the subject's influence on the object. Verbs of this class have a formal expression of the form "noun in the nominative case - verb - noun in the genitive case". Thus, the main semantically significant objects “examiner”, “student”, and the main semantically significant relationship “impact” are identified.

9. The construction of syntactic groups inside the received simple sentences, in which the predicate actants are the main words, with the help of syntactic rules that reveal the syntactic relations between words. The constructed groups are shown in Table 14.

Thus, many other semantically significant objects and attributes are revealed, as well as other semantically related relationships. For this example, they are summarized in Table 15.

After performing the previous steps on the set of selected elementary units of the first, second, third and fourth levels, using the above-mentioned rules, semantically significant relations between semantically significant objects are found. So, for example, in the sentence "A mathematics exam is a written solution to problems" in the text under consideration using a variety of rules, the corresponding signal processing scheme is shown in Figure 2 (processing points 1-9), and the dictionaries used in this rule are presented in Tables 6-16, the semantically significant relation “is” stands out. Other semantically significant relationships are distinguished using the same set of rules. Each semantically significant relation is assigned its type. As a result, semantically significant relationships are distinguished in the source text. Many of these semantically significant relationships with the types assigned to them for the example in question are presented in Table 16.

Thus, after all the above processing steps have been completed, the source text will be marked out with a lot of annotations corresponding to semantically significant objects, attributes and semantically significant relationships between semantically significant objects, as well as between semantically significant objects and attributes.

After that, at the step indicated by reference numeral 29 in FIG. 2, the occurrence frequencies of semantically significant objects and attributes, as well as semantically significant relationships between semantically significant objects and between semantically significant objects and attributes throughout the entire text are detected. This operation is performed in almost the same way as in step 24 for elementary units of the first level. A fragment of such a frequency dictionary for our example is presented in Tables 17 and 18.

The next step, indicated by reference numeral 30 in FIG. 2, is technical and is performed to form triads corresponding to stored semantically significant objects, attributes, and semantically significant relationships. A fragment of the set of such triads for our example is presented in Table 19. In fact, the generated set of triads makes up the initial data for constructing the semantic index processed in the previous stages of the text.

In the step indicated by reference numeral 31 in FIG. 2, a semantic index is constructed as follows: first, from the set of triads obtained in the previous step, subsets of triads are formed, each of which corresponds to one semantically significant object with its attributes, and each obtained subset of triads is used as an input for one of the standard indexers, for example, the widely known Lucene freely distributed indexer, Yandex search engine indexer, Google indexer or any other indexer, from but which are unique to a given subset of triads index. A similar sequence of actions is performed for all subsets of triads corresponding to triads of the form “semantically significant object - semantically significant relation - semantically significant object” and triads of the form “semantically significant object - semantically significant relation - attribute”, receiving a set of corresponding unique indices, which together and make up the semantic index of the text.

In the step indicated by reference numeral 32 in FIG. 2, the indices generated in step 30 of the triad and the indices obtained in step 31, together with a link to the source text from which these triads are generated, are stored in the database.

In accordance with the method of automated semantic comparison of texts in natural language from these triads, a semantic network can be formed in such a way that the first semantically significant object of the subsequent triad is associated with the same second semantically significant object of the previous triad. An example of a fragment of such a semantic network is shown in Table 20.

Moreover, before storing the generated triads and obtained indices in the database, the iterative procedure performs renormalization of the frequencies of occurrence of semantically significant objects and attributes, as well as the frequencies of occurrence of semantically significant relations, into the semantic weight of semantically significant objects and attributes that are the vertices of the semantic network, so that semantically significant objects or attributes connected in the network with a large number of semantically significant objects or attributes with a high frequency occurrences increase their semantic weight, while other semantically significant objects or attributes evenly lose it. An example of the renormalized into semantic weights of the numerical values of the weighting coefficients of the concepts of the semantic network is shown in Table 21. In the same way, samples of texts describing classes (in this example, three) are processed that must be compared with the classified text.

Next, the degrees of intersection of the semantic networks of the classified text and the samples of texts characterizing the classes (subject areas) are calculated, both by vertices and by their relationships, taking into account the semantic weights of the vertices of the semantic networks and the weight characteristics of their relationships. An example of the values of the degrees of intersection of the semantic networks of the classified text and the samples of texts describing the classes (subject areas) are given in Table 22. The degree of intersection of the classified text with the class “Mathematics” indicates their greater semantic similarity compared to other classes.

If the threshold for classifying classified text as subject areas (classes) is set to 2.00000, the text does not fall into any of the specified classes. When setting the threshold to 1.50,000, the text falls into the subject area “Mathematics”.

The degree of intersection of two semantic networks belonging to the classified text and samples of texts describing classes (subject areas) is calculated as the sum of the degrees of intersection of the fifth level elementary units of these two networks. This sum is formed over all the vertices of the network with more vertices. For each vertex of this network, there is a vertex in another network, which is the same elementary unit of the fourth level - the same semantically significant object or the same attribute. If such a vertex is not in the second network, the degree of intersection for this vertex is equal to zero. An example of the values of the degrees of intersection of the vertices of the semantic networks of a classified text and a sample of texts characterizing one of the classes is shown in Table 23.

For each vertex of one semantic network (for each semantically significant element or attribute - elementary units of the fourth level), we calculate the degree of intersection with the corresponding vertex of another semantic network. In the given example, we consider, for example, the vertex “function”, which is available in the semantic networks of both compared texts. This degree of intersection is calculated as the sum of the degrees of intersection of all semantically significant objects and attributes associated with this vertex. In the semantic networks of a classified text and a selection of texts characterizing the "Mathematics" class, these are "equation", "derivative", "score", "equation solution", etc., in one semantic network, and "equation", "derivative", “Solution of an equation”, “order”, etc. - in another semantic network.

For the “function” vertices, the scalar products normalized to 100% are calculated 99 × 99 × sin (52.2 °) / 100 = 77.44 and 99 × 99 × sin (75.6 °) / 100 = 94.93 with the vertices “ the equation". And so for all the vertices of the semantic network, whose semantic weight exceeded the threshold value (chosen equal to 70 in this example).

The total degree of intersection of two semantic networks along the vertex “function” - 177.49 for all vertices of semantic networks adjacent to it is normalized to the largest number of 120 remaining after removal of subthreshold vertices in one of the two semantic networks of the compared texts.

The degree of intersection of semantic networks is thus calculated by summing the smallest degrees of intersection of two pairs of the same semantically significant concepts or attributes of the two compared networks (see Table 24). In this case, semantic intersections of semantic weights of each semantically significant object or attribute associated with this vertex in these two networks are calculated. These semantic intersections are calculated as normalized to 100% scalar products of the semantic weights of the first and second vertices, and the angle between them is taken proportional to the normalized at 100% frequency of occurrence of a semantically significant relationship connecting them. The lesser of the scalar products is added to the amount received. If there is no corresponding semantically significant object or attribute in the second network for a given vertex, the degree of intersection over this semantically significant object or attribute is equal to zero. After summing over all semantically significant objects or attributes associated with the current vertex, the resulting sum is normalized to the largest number of semantically significant objects and attributes associated with this vertex in the two networks and go to the next vertex.

The sum obtained for all vertices in one of the networks (with the largest number of vertices) is normalized to the number of elementary units of the fourth level saved after applying the processing in step 7 (see Fig. 1).

The subject area (class) "Mathematics" is the subject area (class) to which the classified text belongs.

It should be emphasized once again that although the linguistic experts preliminarily construct a lot of syntactic rules and corresponding linguistic dictionaries in the claimed method (as a result of which the definition of “automated” is used in the name of the claimed method), the semantic classification of texts described above is carried out without operator intervention.

Thus, the present invention provides a method for semantic classification of texts in natural language with virtually no operator. The main difference of this method from known methods is that the frequencies of occurrence of elementary units of the fourth level are calculated semantically significant objects and attributes with their subsequent renormalization into semantic weights. Combining triads of semantically significant objects and attributes using semantically significant relationships into a semantic network provides a quick classification of texts, especially texts in highly inflected languages.

Table 1 Token text segmentation Token Number Token Start the end Token Type one In one 2 word 2 to all four 7 word 3 the world 9 12 word four exam fourteen twenty word 5 by 22 24 word 6 math 26 36 word 7 - 38 38 prep sign 8 this is 40 42 word 9 written ... ... word 10 decision ... ... word eleven tasks ... ... word, sentence border 12 . ... ... prep sign 13 Writing ... ... word fourteen character ... ... word fifteen test ... ... word 16 is considered ... ... word 17 everywhere ... ... word eighteen so ... ... word 19 same ... ... word twenty compulsory ... ... word 21 a sign ... ... word 22 a democratic ... ... word 23 of society ... ... word 24 , ... ... prep sign 25 as ... ... word 26 elections ... ... word 27 of ... ... word 28 several ... ... word 29th candidates ... ... word, sentence border thirty . ... ... prep sign 31 Really ... ... introductory word 32 , ... ... prep sign 33 on ... ... word 34 verbally ... ... word 35 exam ... ... word 36 student ... ... word 37 completely ... ... word 38 defenseless ... ... word, sentence border 39 . ... ... prep sign 40 To me ... ... word 41 happened ... ... word 42 hear ... ... word

43 , ... ... prep sign 44 taking ... ... word 45 exams ... ... word 46 on ... ... word 47 the department ... ... word 48 differential ... ... word 49 equations ... ... word fifty mechanical-mathematical ... ... word 51 faculty ... ... word 52 Moscow State University ... ... reduction 53 , ... ... prep sign 54 examiners ... ... word 55 , ... ... prep sign 56 which are ... ... word 57 drowned ... ... word 58 behind ... ... word 59 neighboring ... ... word 60 the table ... ... word 61 students ... ... word 62 , ... ... prep sign 63 giving ... ... word 64 impeccable ... ... word 65 the answers ... ... word 66 (possibly exceeding the level of understanding of the teacher) ... ... plug-in design, supply line 67 . ... ... prep sign 68 Known ... ... word 69 and ... ... word 70 such ... ... word 71 cases ... ... word 72 , ... ... prep sign 73 when ... ... word 74 drowned ... ... word 75 on purpose ... ... word 76 (sometimes this can be saved by entering the audience on time) ... ... plug-in design, supply line 77 . ... ... prep sign

table 2 Lemmas and morphs Token Number Lemmas Morphs one in Pretext 2 all Dat.p Mn. Pronoun Pronoun. Tv.p. M.R. Unit Pronoun Pronoun. Tv.p. Wed Unit Pronoun Pronoun.

Offers M.R. Unit Pronoun Pronoun. Offers Wed Unit Pronoun Pronoun. 3 peace Offers M.R. Unit Noun Inanimate. four exam Named after M.R. Unit Noun Inanimate. Win.p. M.R. Unit Noun Inanimate. 5 by Pretext 6 mathematician Offers M.R. Unit Noun Animated. maths Dat.p J.R. Unit Noun Inanimate. Offers J.R. Unit Noun Inanimate. 8 this is Particle this Named after Wed Unit Pronoun Pronoun. Win.p. Wed Unit The pronoun Inanimate. Pronoun Win.p. Wed Unit Pronoun Animated. Pronoun 9 writing Win.p. Wed Unit Adjective Animation. Named after Wed Unit Adjective Animation. 10 decision Named after Wed Unit Noun Inanimate. Win.p. Wed Unit Noun Inanimate. eleven a task Rod.p. J.R. Mn. Noun Inanimate. 13 writing Win.p. M.R. Unit Adjective Animation. Named after M.R. Unit Adjective Animation. fourteen character Named after M.R. Unit Noun Inanimate. Win.p. M.R. Unit Noun Inanimate. fifteen test Rod.p. Wed Mn. Noun Inanimate. 16 be considered Unit Real 3rd person Verb Imperfect 17 everywhere Adverb eighteen so Adverb 19 same Particle twenty required Dat.p Mn. Adjective Tv.p. M.R. Unit Adjective Tv.p. Wed Unit Adjective 21 sign Tv.p. M.R. Unit Noun Inanimate. 22 democratic Rod.p. M.R. Unit Adjective Rod.p. Wed Unit Adjective Win.p. M.R. Unit Adjective Animation. Win.p. Wed Unit Adjective Animation. 24 society Named after Wed Mn. Noun Inanimate. Rod.p. Wed Unit Noun Inanimate. Win.p. Wed Mn. Noun Inanimate. 25 as Union 26 the choice Named after M.R. Mn. Noun Inanimate. Win.p. M.R. Mn. Noun Inanimate 27 of Pretext 28 some Rod.p. Mn. Numeric Quantitative Offers Mn. Numeric Quantitative Win.p. Mn. Numerical Quantitative Animation. 29th candidate Rod.p. M.R. Mn. Noun Animated. Win.p. M.R. Mn. Noun Animated. 35 on Pretext 34 oral Offers M.R. Unit Adjective Offers Wed Unit Adjective 35 exam Offers M.R. Unit Noun Inanimate.

36 student Named after M.R. Unit Noun Animated. 37 completely Adverb 38 defenseless M.R. Unit Short f. Adjective 40 I am Dat.p Unit 1st person Pronoun Personal (odush) Offers Unit 1st person Pronoun Personal (odush) 41 happen Wed Unit Past Verb Imperfect 42 hear Verb Imperfect 44 accept Real Holy Communion Imperfect 45 exam Named after M.R. Mn. Noun Inanimate. Win.p. M.R. Mn. Noun Inanimate. 46 on Pretext 47 the department Dat.p J.R. Unit Noun Inanimate. Offers J.R. Unit Noun Inanimate. 48 differential Rod.p. Mn. Adjective Win.p. Mn. Adjective Animation. Offers Mn. Adjective 49 the equation Rod.p. Wed Mn. Noun Inanimate. fifty mechanical-mathematical Rod.p. M.R. Unit Adjective Rod.p. Wed Unit Adjective 51 faculty Rod.p. M.R. Unit Noun Inanimate. 54 examiner Rod.p. M.R. Mn. Noun Animated. Win.p. M.R. Mn. Noun Animated. 56 which the Named after Mn. Pronoun Pronoun. Win.p. Mn. The pronoun Inanimate. Pronoun 57 to drown Mn. Past Verb Imperfect 58 behind Pretext 59 neighboring Dat.p Mn. Adjective Tv.p. M.R. Unit Adjective Tv.p. Wed Unit Adjective 60 table Tv.p. M.R. Unit Noun Inanimate. 61 student Rod.p. M.R. Mn. Noun Animated. Win.p. M.R. Mn. Noun Animated. 63 to give Rod.p. Mn. Past Active Communion Imperfect Win.p. Mn. Past Active Communion Animated. Imperfect Offers Mn. Past Active Communion Imperfect 64 impeccable Win.p. Mn. Adjective Inanimate. Named after Mn. Adjective Inanimate. Named after M.R. Mn. Noun Inanimate. 65 answer Win.p. M.R. Mn. Noun Inanimate. 68 famous Mn. Short f. Adjective 69 and Union 70 such Named after Mn. Pronoun Pronoun. Win.p. Mn. The pronoun Inanimate. Pronoun Named after M.R. Mn. Noun Inanimate. 71 happening Win.p. M.R. Mn. Noun Inanimate. 73 when Union 74 drowned Mn. Past Verb Imperfect 75 on purpose Adverb

Table 3 Frequencies of occurrence of the first and subsequent words of the sequence in the text, as well as differences in the frequencies of occurrence for different words in the sequence Repeating a sequence of words in a text Sequence words Frequency of occurrence Frequency difference one one asymptotically one stable one 0 2 asymptotically 2 stable 2 0 3 asymptotically 3 stable 3 0 ... ... asymptotically 7 0 7 stable 7 0 ... ... ... ... ...

Table 4 Stable word combinations in the text Collocation asymptotically stable ... ...

Table 5 Lots of text suggestions Offer No. Text suggestions Level 1 Units Level 2 Units Level 3 Units one Throughout the world, a math exam is a written solution to problems. In, all, the world, exam, in, mathematics, this, written, solution, tasks all, world, exam, by, (mathematician, mathematics), (this, this) written, solution, problem 2 The written nature of the trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates. Written, character, trials, is considered, everywhere, to be the same, obligatory, sign, democratic, society, as, election, of, several, candidates written, character, test, reckoned, everywhere, as, same, mandatory, sign, democratic, society, like, choice, of, several, candidate 3 Indeed, in an oral exam, a student is completely defenseless. On, oral, exam, student, completely, defenseless on, oral, exam, student, totally, defenseless four I happened to hear taking exams at the Department of Differential Equations of Mechanical and Mathematical I happened to hear, taking, exams, at the department, differential, equations, mechanical and mathematical, i, happen, hear, take, exam, on, chair, differential, equation,

Faculty of Moscow State University, examiners who drowned students at the next table who gave impeccable answers (possibly exceeding the level of understanding of the teacher). Faculty, examiners who, drowned, at, the next, table, students who gave, impeccable answers mechanics and mathematics, faculty, examiner, who, drowning, at, neighboring, desk, student, give, impeccable, answer 5 There are also known cases when they drowned on purpose (sometimes it can be saved from this by entering the audience on time). There are known, and such cases, when they drowned, on purpose known, and such a case, when, to drown, on purpose

Table 6 Fragment Type Personal verb Brief communion Short adjective Predicative word Participle Communion Infinitive Introductory word Other one 2 3 four 5 6 7 8 9

Table 7 Initial Fragmentation Results No. of fragments. Offer snippets Fragment Type 1.1 math exam worldwide is a written solution to problems DASH 2.1 the written nature of trials is considered everywhere an equally indispensable feature of a democratic society one 2.2 as an election of several candidates 9 3.1 on an oral exam, the student is completely defenseless 3 4.1 I happened to hear one 4.2 taking exams at the department of differential equations of the faculty of mechanics and mathematics 6 4.3 examiners 9 4.4 who drowned students at the next table one 4.5 giving perfect answers 5 5.1 are known 3 5.2 and such cases 9 5.3 when drowned on purpose one

Table 8 Description attribute level elements Offer Components Morphological features Object / Subject Noun pronoun Act Verb Item Feature Full adjective, ordinal, pronoun-adjective, consistent with the object / subject by gender, number and case Sign of action Adverb Attribute measure Adverb, adverbial numeral

Table 9 Syntactic groups corresponding to the attribute level of description Offer No. Syntax Group Elements Token Numbers Syntax group one feature of the object + object 2 + 3 all over the world one feature of the object + object 9 + 10 written decision 2 feature of the object + object 13 + 14 written character 2 action + sign of action 16 + 17 considered everywhere 2 sign measure + sign of the object + object 18 + 19 + 20 + 21 an equally mandatory attribute 2 feature of the object + object 22 + 23 a democratic society 2 feature of the object + object 29 + 30 several candidates 3 feature of the object + object 35 + 36 oral exam 3 feature measure + feature feature 38 + 39 completely defenseless four feature of the object + object 49 + 50 differential equations four feature of the object + object 51 + 52 Faculty of Mechanics and Mathematics four feature of the object + object 60 + 61 next table four feature of the object + object 65 + 66 immaculate answers 5 feature of the object + object 71 + 72 such cases 5 action + sign of action 75 + 76 drowned on purpose

Table 10 Received simple sentences as a result of enlargement of fragments No simple offer Enlarged Fragments Simple sentences 1.1 math exam worldwide is a written solution to problems math exam worldwide is a written solution to problems 2.1 the written nature of trials is considered everywhere an equally indispensable feature of a democratic society the written nature of trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates as an election of several candidates 3.1 on an oral exam, the student is completely defenseless on an oral exam, the student is completely defenseless 4.1 I happened to hear taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics I happened to hear taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics 4.2 examiners stoked students at the next table, giving impeccable answers examiners stoked students at the next table, giving impeccable answers 5.1 such cases are known such cases are known 5.2 when drowned on purpose when drowned on purpose Note to the table: the first digit in the number of the simple proposal corresponds to the number of the proposal to which it belongs.

Table 11 Minimum block diagrams of sentences (fragment) MCC Examples of offers N1 V (f) The Rooks Have Arrived. Things are done by people. N1 Cop (f) Adj1 The night was quiet (quiet, quiet). N1 Cop (f) Adj5 The night is quiet (quiet). N1 Cop (f) Adj (f) The night was quieter than the day. N1 Cov (f) N1 He (was) a student. N1 Cop (f) N5 He was a student. Cop (f) N1 It will be raining. It was winter. Whisper. Timid breathing. Silence. ...

Explanation of table 11:

V (f) - conjugated forms of the verb (not infinitive);

Cop (f) - conjugated forms of a bunch of official words to be, become, appear;

Inf - infinitive of a verb or connective;

N1, N5 - nominative, instrumental case of the substantive;

Adj1, Adj5 - nominative, instrumental case of adjectives and passive participles;

Adj (f) - short forms and corporate bodies of adjectives and passive participles.

Sentences with the pattern Cop (f) N1 can be called, i.e. the link verb is not present there explicitly. In this case, we assume that the predicate is zero, denoted as NULL.

Table 12 The predictive minimum of simple sentences that make up the complex sentence of the source text No simple offer Simple sentences MCC Template Predictive minimum (Subject-Predicate) 1.1 math exam worldwide is a written solution to problems N1 Cop (f) N1 Substantive in the nominative case + Copula + Substantive in the nominative case (exam; there is a solution) 2.1 the written nature of trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates N1 Cop (f) N1 / 5 N2 - Substantive in the nominative case + Copula + Substantive in the instrumental case (character, election; considered a sign) 3.1 on an oral exam, the student is completely defenseless N1 Cop (f) Adj1 / 5 Substantive in the nominative case + Copula + short adjective in the nominative case (student; there is defenseless) 4.1 I happened to hear, taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics, examiners N3 V (f) Inf Substantive in the dative case + conjugated form of the verb + infinitive (me; happened to hear) 4.2 examiners stoked students at the next table, giving impeccable answers N1 V (f) Substantive in the nominative case + conjugated form of the verb (examiners; drowned) 5.1 such cases are known N1 Cop (f) Adj1 / 5 Substantive in the nominative case + Copula + short adjective in the nominative case (cases; there are known) 5.2 when drowned on purpose Conjugate Plural (NULL; stoked)

Table 13 Fill valencies for example text predicates No simple offer Predicate 1. Subject 2. Object 3. Addressee 4. Tool 5-7. Locatives 1.1 there is a solution exam tasks - - - 2.1 considered a sign character election - - - - 3.1 is defenseless student - - - - 4.1 happened to hear to me examiners - - - 4.2 drowned examiners students - - - 5.1 have known cases - - - - 5.2 drowned - - - - - Note to the table: 5 - the initial locative, 6 - the final locative, 7 - the middle locative.

Table 14 Syntactic groups derived from source text using syntax rules No simple offer Simple sentences Syntactic groups where actants and a predicate are the main words Name of groups and rules 1.1 math exam worldwide is a written solution to problems math exam Genitive determination in postposition written decision Object + Object Tag problem solving Genitive determination in postposition 2.1 the written nature of trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates written character Object + Object Tag nature of testing Genitive determination in postposition a mandatory feature Object + Object Tag sign of society Genitive determination in postposition

election of candidates Genitive determination in postposition 3.1 on an oral exam, the student is completely defenseless - - 4.1 I happened to hear, taking exams at the Department of Differential Equations of the Faculty of Mechanics and Mathematics, examiners - - 4.2 examiners stoked students at the next table, giving impeccable answers impeccable students Object + Object Tag 5.1 such cases are known such cases Object + Object Tag 5.2 when drowned on purpose - -

Table 15 Many semantically significant objects and attributes (fragment) Simple sentence Semantically significant objects Attributes math exam worldwide is a written solution to problems exam mathematics decision written tasks the written nature of trials is everywhere considered as an indispensable feature of a democratic society as the election of several candidates character writing test sign required of society elections from candidates

Table 16 Relations between semantically significant objects, and between semantically significant objects and attributes Semantically significant object 1 Semantically significant object 2 Semantically significant relationship Type of semantically significant relationship one exam decision there is be 2 character sign be considered be 3 examiners students to drown to influence ... ... ... ... ...

Table 17 The frequency of occurrence of semantically significant objects and attributes Semantically significant object or attribute Frequency of occurrence one teacher fourteen 2 student 27 3 function 16 four the equation 44 ...

Table 18 Frequency of occurrence of semantically significant relations between semantically significant objects and between semantically significant objects and attributes Semantically significant object 1 - semantically significant object 2 Semantically significant relationship The frequency of the semantically significant relationship one student teacher change state 8 2 function - equation why four ...

Table 19 Many triads (fragment). Triads one exam - solution 2 teacher - student 3 student exam ...

Table 20 Semantic network of triads (fragment). Main word Attitude Subordinate Word one teacher change state student 2 student be evaluated exam 3 exam be decision ...

Table 21 The semantic weight of semantically significant words and attributes Semantically significant object or attribute Semantic weight one teacher 99 2 student 99 3 function one hundred four the equation 99 ...

Table 22 The degrees of intersection of semantic networks of the source text with the networks of two other texts Classes Maths Education Story Degree of intersection 1,59160 0.46480 0.18382

Table 23 The degree of intersection at one vertex of the semantic network of the classified text and the semantic network of domain text selection fragment of the first network fragment of the second network degree of intersection of second objects or attributes vertex vertex associated with the first semantic weight, relationship weight vertex associated with the first semantic weight, relationship weight function 177.49 / 120 = 1.4791 one the equation 100, 58 the equation 99.84 77.44 2 derivative 99.48 derivative 99.62 67.09 8 point 99.48 0 3 equation solution 99.32 equation solution 87.25 32.96 order 99.62 0 argument 97.57 0 degree of intersection of vertices "function" 177.49 ...

Table 24 The degree of intersection of the semantic networks of the classified text and the selection of texts of one of the subject areas first network second network vertex vertex associated with the first vertex vertex associated with the first total weight one the equation one the equation 14.25 2 function 2 function 15.15 3 point 3 argument 0 four plane four plane 13.10 6 derivative 6 derivative 16.23 8 decision 8 decision 15,20 9 point 9 point 14.01 ... 76 vector field 76 process 0 625 differential equation 0 Amount 994.75 Normalized amount 994.75 / 625 = 1.5916

Claims (6)

1. The method of automated semantic classification of texts in natural language, which consists in the fact that:
- present each classified text in digital form for subsequent automatic and (or) automated processing;
- carry out the indexation of each classified text in digital form, receiving:
- elementary units of the first level, including at least words,
- elementary units of the second level, each of which is a normalized word form,
- elementary units of the third level, each of which is a stable phrase in the said text,
- elementary units of the fourth level, each of which is a semantically significant object and attribute, and
- elementary units of the fifth level, each of which is a triad of either two semantically significant objects and a semantically significant relationship between them, or from a semantically significant object and attribute and a semantically meaningful relation connecting them;
- identify the frequency of occurrence of elementary units of the fourth level and the frequency of occurrence of the mentioned semantically significant relationships;
- store in the database the formed elementary units of the second, third, fourth and fifth levels with the identified frequencies of occurrence of elementary units of the fourth level and semantically significant relationships, as well as the resulting indices, together with links to specific sentences of this text;
- form a semantic network from said triads in such a way that the first elementary unit of the fourth level of the subsequent triad is associated with the same second elementary unit of the fourth level of the previous triad;
- carry out, in the course of an iterative procedure, renormalization of the mentioned frequencies of occurrence into the semantic weight of elementary units of the fourth level, which are the vertices of the semantic network, so that elementary units of the fourth level connected in the network with a large number of other elementary units of the fourth level with a high frequency of occurrence, increase their semantic weight, and other elementary units of the fourth level evenly lose it;
- rank elementary units of the fourth level by semantic weight by comparing the semantic weight of each of them with a predetermined threshold value and delete elementary units of the fourth level having semantic weight below the threshold value;
- retain in memory the remaining elementary units of the fourth level with a semantic weight above the threshold, as well as semantically significant relations between the remaining elementary units of the fourth level;
- reveal the degree of intersection of the mentioned semantic network of classified text and semantic networks of text samples, which text samples are composed of previously classified texts and describe the subject areas of the mentioned semantic classification, while the said degree of intersection is revealed both by the vertices of the mentioned semantic networks, and by the relationships between these vertices, taking into account the semantic weights of the vertices of the considered semantic networks and the weight characteristics of their relationships, and take the identified stump semantic nets crossing the classified text and text specific sample as the quantity characterizing the semantic similarity of the classified text and this text sample;
- choose, as a class for the classified text, at least one of the mentioned subject areas, the degree of intersection of the semantic network of which with the semantic network of said classified text are greater than a predetermined threshold. 2. The method according to claim 1, in which, when said excess of said predetermined threshold is exceeded by degrees of intersection for several subject areas, said subject areas are ranked by their degree of proximity to the classified text. 3. The method according to claim 1 or 2, in which a predetermined number of the aforementioned subject areas to which the classified text is referred is selected. 4. The method according to claim 1, wherein said indexing is carried out in the process of performing the following steps:
- segment the text in digital form into elementary units of the first level, including at least words;
- segment text in digital form into sentences according to graphematical rules;
- form for each elementary unit of the first level, which is a word, on the basis of morphological analysis, elementary units of the second level, including a normalized word form;
- calculate the frequency of occurrence of each elementary unit of the first level for two or more adjacent units of the first level in this text and combine among the mentioned elementary units of the first level sequences of words following one after another in this text into elementary units of the third level, which are stable combinations of words , if for every two or more consecutive words in a given text the difference is in the calculated frequencies of occurrence of these words for the first occurrence of a given sequence telnosti words and a number of subsequent occurrences for each pair of sequences of words remain unchanged;
- identify, in the process of multi-stage semantic-syntactic analysis by referring to linguistic and heuristic rules pre-generated in the database in a predefined linguistic environment, in each of the generated sentences, semantically significant objects and attributes are units of the fourth level;
- for each elementary unit of the fourth level, the identity is fixed by reference between the corresponding semantically significant object, as well as the attribute, and the corresponding anaphoric reference, if any, in the classified text, replacing each anaphoric reference with the corresponding antecedent;
- store in memory each semantically significant object and attribute;
- identify, in the process of multistage semantic-syntactic analysis by referring to linguistic and heuristic rules pre-generated in the database in a predefined linguistic environment, in each of the generated sentences, semantically significant relationships between the identified units of the fourth level - semantically significant objects, as well as between semantically significant objects and their attributes;
- assign to each semantically significant relation the corresponding type from the subject ontology stored in the database on the subject of the subject area to which the classified text belongs;
- identify throughout the text the frequency of occurrence of elementary units of the fourth level and the frequency of occurrence of the mentioned semantically significant relationships;
- store in memory each identified semantically significant relation together with the type assigned to it;
- form within the given text for each of the identified semantically significant relations that connect both the corresponding semantically significant objects and the semantically significant object and its attribute, a multitude of triads, which are elementary units of the fifth level;
- index on the set of formed triads individually all related semantically meaningful relations semantically significant objects with their frequencies of occurrence, all attributes with their frequencies of occurrence, and all formed triads. 5. The method according to claim 1, in which the said degree of intersection of two semantic networks is calculated as the sum of the matches of the elementary units of the fifth level of these two semantic networks. 6. The method according to claim 5, in which:
- choose one of the two semantic networks as the basic network, in which, after ranking and removing vertices with semantic weights, more vertices remain below the threshold value than the other, which is chosen as the comparison;
- find for each vertex of said core network in said comparative network a vertex that is the same elementary unit of the fourth level, i.e. the same semantically significant object, or the same attribute;
- calculate, for each vertex found in each of the aforementioned base and compared networks, the values of all the triads referred to this vertex as the area of triangles, the sides of which correspond to the components of each of these triads, and the angle between the sides is proportional to the weight of the semantically significant relation;
- choose for each pair of said triads associated with a pair of specific vertices in said basic and compared networks, the smaller of said calculated values as the degree of intersection of said triads in said basic and compared networks;
- summarize for each vertex associated with a given vertex all selected calculated values, obtaining the degree of intersection for a given pair of vertices of the mentioned base and compared networks;
- normalize the found amount to the number of the mentioned semantically significant objects and attributes associated with a given vertex in that of the aforementioned base and compared networks, which contains more vertices associated with a given vertex;
- summarize the normalized sums for all the vertices of that of the mentioned base and compared networks, which contains more vertices;
- normalize the received amount to the number of elementary units of the fourth level remaining in this network, obtaining the mentioned degree of intersection of two semantic networks.

Images