KR102065083B1 - An English sentence translation system with machine learning-based direct-reading symbols - Google Patents

Abstract

The present invention relates to an English sentence translation providing system with direct-reading symbols based on machine learning. The English sentence translation providing system with direct-reading symbols based on machine learning according to the present invention comprises: a content input part receiving English content from a user terminal; a direct-reading symbol generation part adding a direct-reading symbol to the English content received in the content input part; a translation part receiving an English sentence including the direct-reading symbol generated in the direct-reading symbol generation part and performing direct-reading translation to generate a direct-reading translation; and a content transmission part providing the English content with the direct-reading symbol generated from the direct-reading symbol generation part and the direct-reading translation to the user terminal.

Description

An English sentence translation system with machine learning-based direct-reading symbols}

The present invention relates to a system for providing English sentence translation based on direct learning symbols based on machine learning.

So far, English-Korean machine translation systems usually consist of three stages: English interpretation, translation, and Korean generation. In particular, the English interpretation stage, which determines the performance of the machine translation system, goes through three processes: English morphological analysis, syntax analysis, and meaning generation.

In general, the parts of speech are designated in the morphological analysis process and the parts of speech are analyzed in the syntax analysis process. The success rate of the sentence in Chinese or compound sentence is lower than that of the short sentence, and the more complicated and longer the sentence is, like the mixed sentence. There is a high probability of failure.

If you try to analyze the whole sentence at once, there is often an error in the lack of rules or the ambiguity in the syntax.

As a solution to this problem, the long sentence is divided into short paragraphs to analyze the structure of the English sentence. Its role can be clarified in terms of clauses, and its lower ambiguity in word units allows for a more accurate translation.

Although such efforts have improved the existing English-language translators with fewer errors and shorter translation times, there are still some disadvantages. Most of these systems do not escape their structural limitations. In other words, in case of long sentences, it is necessary to include a sectioning process to reduce the failure of structural analysis, and in order to determine one sentence component within a section unit, all possible forms are sequentially performed from word to phrase and clause level. Candidates should be verified. You must repeat the same process for each sentence component and then analyze the syntax by applying sentence rules.

This is an unavoidable phenomenon because the parsing, the backbone of the existing translator, is based on the existing English grammar system. Sentence elements are given as grammar elements, and they are divided into verbs, verbs, objects, bores, and formulas (place, time, reason, method), and the sixth principle is added to the English text (who, what, where, when, why, how). , It is to classify sentence patterns based on it.

In order to interpret English sentences based on this grammar, first, the clauses must be separated, and then the sentence patterns on the truncation must be classified and given meaning according to the sentence components. As many grammar rules as there are base sentences are involved, and their application is repeated from clause to clause. In addition to the modified rules, the interpretation process becomes long and complicated.

In addition, even after the parsing is completed with the designation of sentence components, the detailed structure must be reviewed as a whole in the semantic step. Otherwise, it is also a serious flaw that semantic processing is often impossible or incomplete. This is also a major cause of translation errors. In particular, since the complex structure of verb phrases developed in English cannot be analyzed in detail, accurate meaning cannot be obtained unless the original text is reviewed and moved when assigning meaning.

Patent 10-1302875

The present invention has been made to solve the above problems and by analyzing the English sentence by sentence component or part of speech and sentence data by the symbol specific to each sentence component or part of speech, it is possible to quickly and easily grasp the sentence structure and identified It is an object of the present invention to provide a system for providing an English sentence translation system having a direct reading symbol based on machine learning, which enables rapid and accurate translation by proceeding translation using a sentence structure.

The present invention provides a content input unit for receiving English content from a user terminal; A direct reading symbol generator for adding a direct reading symbol to the English content inputted from the content input unit; A translation unit configured to receive an English sentence including a direct reading symbol generated by the direct reading symbol generator and perform a direct reading to generate a direct reading translation; And a content transmission unit providing the user terminal with a direct reading translation and the English content added with the direct reading symbol generated from the direct reading symbol generator.

According to the present invention, by analyzing the English sentence by sentence component or part-of-speech and displaying sentence data with symbols specific for each sentence component or part-of-speech, it is possible to quickly and easily grasp the sentence structure and to proceed with the translation using the identified sentence structure. Ensure fast and accurate translations.

1 is a block diagram of an English sentence translation providing system having a direct reading based on machine learning proposed by the present invention.
2 is an exemplary view of sentence data on which symbol display is performed according to the present invention.
3 is an exemplary diagram in which a translation is directly read.
4 is an exemplary view in which a Korean word order translation is provided.
5 is a diagram illustrating a conceptual image generation process.
6 is a view illustrating a comparison image generation process.
7 is a detailed block diagram of the translation unit of FIG. 1.

Further features and effects of the present invention will be clearly understood from the best mode for carrying out the invention described below.

Example

EMBODIMENT OF THE INVENTION Hereinafter, this invention is demonstrated in detail according to a preferable embodiment, referring an accompanying drawing.

1 is a block diagram of an English sentence translation providing system having a direct reading based on machine learning proposed by the present invention.

Referring to FIG. 1, the English sentence translation providing system 1 having a machine learning-based direct reading symbol proposed by the present invention includes a content input unit 2, a direct reading symbol generating unit 3, and a translation unit 3-. 1) it may be divided into a concept image generator 3-2, a comparison image generator 3-3, and a content transmitter 4;

The content input unit 2 provides a function of receiving English content such as text, an image, and a video from a user terminal. Herein, the English content means content including an English sentence, and the content input unit 2 extracts an English sentence and provides it directly to the symbol generator 3 when the text is input, and in the case of an image or a video. The English sentence included in the image or video is extracted and directly provided to the symbol generator 3 to extract the English sentence.

The content input unit 2 includes a mobile phone camera, and may receive an English sentence photographed through the mobile phone camera, or may receive an English sentence from the web.

Next, the direct reading symbol generator 3 detects the context based on the English content received through the content input unit 2 and adds the most suitable direct reading symbol.

In addition, the translation unit 3-1 receives an English sentence containing a direct reading symbol generated by the direct reading symbol generating unit 3 and performs a direct reading and a Korean word order translation to perform a direct reading translation and a Korean word order translation. Create

Here, the translation unit 3-1 performs a semantic unit operation of the direct reading symbol in the English sentence with the direct reading symbol, and performs the direct reading of the semantic unit based on the reading.

Next, the concept image generator 3-2 directly generates a concept image according to translation.

Here, the concept image refers to an image produced so that the concept is sequentially reminded according to the translation.

In addition, the concept image generation unit 3-2 may additionally generate and provide a direct explanation symbol description image, a claim (negative) meaning unit display image, a conjunction verb display image, a negative phrase display image, and the like.

The comparison image generating unit 3-3 generates a comparison image, which is made so that concepts are sequentially associated with the Korean word order translation.

In addition, the content transmitter 4 provides the user terminal with the content, the direct translation, the Korean word order translation, the concept image, and the comparison image to which the direct reading symbol is added from the direct reading symbol generating unit 3.

 The English sentence translation providing system 1 having a machine learning-based direct reading symbol and the user terminal may exchange data with each other using near field communication or long distance communication.

The short-range communication applied here may include Bluetooth, Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), ZigBee, and Wireless Fidelity (Wi-Fi) technologies.

In addition, applicable telecommunications include code division multiple access (CDMA), frequency division multiple access (FDMA), time division multiple access (TDMA), orthogonal frequency division multiple access (OFDMA), and single carrier frequency division multiple access (SC-FDMA). Technology).

The direct reading symbol generator 3, the translator 3-1, the concept image generator 3-2, and the comparison image generator 3-3 may be implemented based on machine learning. Content input unit (2), direct reading symbol generator (3), translator (3-1), conceptual image generator (1) constituting an English sentence translation system having a direct reading symbol based on a machine learning ( 3-2), specific functions of the present invention will be described based on the elements of the comparison image generator 3-3 and the content transmitter 4.

The content input unit 2 according to the present invention receives English content such as text, voice, image, and video from a user terminal.

Subsequently, the direct reading symbol generator 3 analyzes the morpheme based on the input English content.

In more detail, the direct reading symbol generator 3 analyzes a sentence component consisting of two or more combinations of a predicate, a target word, a bore word, and other formula words, and analyzes the sentence form of any one of the first to fifth forms. And the first result data of the sentence component and sentence form is generated accordingly.

In addition, the direct reading symbol generator 3 receives the first result data in which the sentence component and sentence form of the sentence data are analyzed, and the noun, the verb, the modifier, the infinitive, the same noun, the participle, the comparative class, the conjunctional conjunction, the conjunction and the relative pronoun. The second result data classifying the sentence data by parts of speech is performed by analyzing the sentence components using any one or a combination of parts of speech parts including a relative adverb.

In addition, the direct reading symbol generator 3 receives the first result data and the second result data to generate a symbol processing sentence data by displaying a specific symbol for each part-of-speech in sentence data, and the specific symbol is a corresponding word or phrase. Underlined or displayed in the original letters, triangles, arrows, etc., the symbol according to an embodiment of the present invention is only one example, such symbol display may be variously adopted and displayed.

In addition, the content transmission unit 4 receives the symbol processing sentence data from the direct reading symbol generation unit 3 and provides it to the user terminal.

2 is an exemplary view of sentence data on which symbol display is performed according to the present invention.

Referring to the drawings, according to an embodiment of the present invention, the direct symbol generator 3 performs a specific symbol processing by classifying by sentence component or part of speech, and in detail, the classified sentence component or part of speech is a verb. If it is, underline the bottom of the word or phrase.

In the case of modifiers (prepositions + nouns), the symbol '/' is displayed on the front side of the word or phrase.In the case of infinitive, the 'to' part accepts the circle and the 'to' part is underlined on the lower part. Draw it.

' 기호를 표시한다.In the case of the injection phrase (V · ing noun (S), pp noun), as mentioned above, the word or phrase above the word or phrase is not processed without the underscore and the symbol of the circle.

'Sign.

In addition, if the classified sentence component or part of speech is a parting (V · ing noun, pp), the arrow on the upper side in the noun direction of the front side or the rear side that the word or phrase modifies in the word V · ing, pp) or phrase. Draw to display.

In addition, if the classified sentence component or part of speech is comparative, the '△' symbol is displayed in the 'as' or 'than' part of the word or phrase.If the classified sentence component or part of speech is an conjunction, the word or phrase '△' or '○' symbol is accepted in order to display this. If the conjunction is an equivalent conjunction, the symbol '△' is displayed.

'기호를 함께 표시한다.In addition, if a classified sentence component or part of speech is a relative pronoun or a relative adverb, the symbol '○' is accepted in the word or phrase, and the forward pronoun or relative adverb

'Together with the symbol.

'기호를 표시한다.In addition, if it is determined that the conjunction, relative pronoun, or relative adverb is omitted through the first result data and the second result data, the symbol '∨' is indicated in the omitted position of the sentence data and the relative pronoun or relation adverb is displayed. If

'Mark the symbol.

Meanwhile, the translator 3-1 divides the sentence by extracting all possible split point candidates while reading the input sentence from the beginning to the end. Here, the split point candidates used are punctuation marks, modifier phrases, and modifier clauses.

Here, punctuation marks include periods (.), Question marks (?), Exclamation points (!), Commas (,), suffixes (·), dashes (:), semicolons (;), dashes (/), and double quotation marks ( “”), Single quotes (''), parentheses (()), braces, ({}), brackets ([]), ellipses (-), ellipses (-), tildes (~), ellipses (..) ....).

At this time, the modifier or modifier clause is directly read by the symbol '/', '△' or '○'.

However, if the noun or noun clause is continued in the case of a comma, the translation unit 3-1 excludes the split point candidate. At this time, the translation unit 3-1 does not exclude the split point candidate when the word following the comma is a preposition.

In the case of the modifier phrase, the translation unit 3-1 excludes the split point candidate when the preceding word is the injection phrase.

' 기호가 표시되어 있는 경우에 분할점 후보에서 제외한다. 다만 번역부(3-1)는 '/' 기호의 뒤에 전치사를 앞에 둔 동명사가 위치하는 경우(/전치사+동명사)에는 분할점 후보에서 제외하지 않는다. That is, in the case of the '/' symbol, the translator 3-1 has a word in front of the spray phrase '

If the symbol 'is displayed, it is excluded from the split point candidate. However, the translation unit 3-1 does not exclude the split point candidate when the same name preceded by the preposition is placed after the / symbol.

In addition, the translation unit 3-1 adds a virtual split point before and after the verb to split the verb.

That is, the translation part 3-1 adds a virtual split point before and after the underline in the case of the verb.

The translator 3-1 excludes the split point candidate when the verb is a verb, that is, a noun.

However, the translation unit 3-1 adds a virtual split point before and after the but before the verb in the conjunctive conjunction.

That is, the translation unit 3-1 is denoted by a '△' symbol in the case of an equivalence conjunction, and is excluded from the candidate for split point when there is no underscore symbol representing a verb after the '△' symbol. In the case of underscores, virtual split points are added before and after.

The translator 3-1 translates a limited part defined by a relative pronoun or a relation adverb by replacing a verb of the modification clause when the expression clause is a relative pronoun or a relation adverb.

'기호를 함께 표시되어 있는바, '○'기호에 '

'기호가 있는 경우에 한정 부분으로 수식어절의 동사를 대체하여 번역한다. 도 3은 이와 같이 직독직해 번역문이 제공되는 예시도이다.That is, the translator 3-1 displays a symbol "○" when a modifier clause is a related pronoun or a relative adverb and moves forward in the direction of a relative pronoun or a relative adverb.

'The symbol is marked with' ○ '

'If there is a symbol, translate it by replacing the verb of the qualifier clause with the limited part. 3 is an exemplary view in which a direct translation is provided in this way.

On the other hand, the translation unit (3-1) not only provides a direct reading translation but also provides a translation suitable for the Korean word order.

At this time, the translation unit 3-1 may be reversed in the case of the modifier associated with the subject.

For example, if the English sentence is "On a clear day, a crewmember on a merchant ship sailing across the caribbean Sea peers out at the horizon through his telescope," on a clear day, a crew member flies his telescope on a merchant ship across the Caribbean. Stare at the horizon through the words.

To prevent this mistranslation, the translator (3-1) extracts nouns before the verb.

At this time, the translation unit 3-1 extracts nouns with strong semantic elements, and the translation unit 3-1 selects a crew member and a merchant ship as an example. As such, the translation unit 3-1 performs the following operation when there are at least two nouns before the verb.

The translator 3-1 inputs a crew member and a merchant ship from the big data, extracts sentences including the corresponding word from the big data, calculates a sentence ratio in which the merchant ship is represented after the crew member in the extracted sentences, and On the contrary, after the merchant ship, the crew calculates the sentence ratio expressed and prepares the Korean word order translations arranged in the order of the largest sentence ratio, and provides the user with the content transmission unit 4.

The translator 3-1 has a 30% sentence ratio in which the merchant ship is expressed after the crew in the sentences extracted as an example, and the sentence ratio calculated by calculating the sentence ratio in which the crew member is represented after the merchant ship is 70%. In percent, the translation unit (3-1) produces a Korean word order translation, translated, "On a clear day, a crew stares at the horizon through his telescope on a merchant ship across the Caribbean."

That is, the translation unit 3-1 reads the semantic unit of the modifier after reading the semantic unit of the subject through the above-described process, and makes the semantic unit of the verb read so that the sentence becomes a natural sentence.

Meanwhile, the translation unit 3-1 performs a post-processing of verbs to correct translational errors with respect to verbs.

To this end, the translation unit 3-1 extracts search nouns before and after the verb and searches for and extracts sentences including related search nouns from the big data.

The translator 3-1 extracts a verb expression having a meaning similar to that of the corresponding verb as a verb replacement candidate, calculates a sentence ratio with respect to the extracted verb replacement candidates, and then has the largest verb with the largest sentence ratio. Replace the verb expression with the candidate for substitution.

For example, the translator 3-1 primarily translates "A black flag is flying high on its mast." To "A black flag is high on its mast." After extracting the as a plurality of representative nouns of the first translation, search for this in the big data to collect sentences including the plurality of representative nouns.

At this time, the sentences that can be collected by the translation unit 3-1 are, for example, "flags of torn flags in the masts dance sadly", "flags among the towering masts", " At the end of the broken mast there was a terribly torn flag hanging. ”The translator (3-1) extracts a verb expression with a similar meaning to the verb's meaning as a verb replacement candidate and calculates a sentence ratio.

As a result of calculating the sentence ratio, the translator 3-1, for example, substitutes "dancing sad dance", which is an alternative verb candidate extracted from "flags dancing in the mast, torn apart flags." Includes 2% sentence rate, including the alternative verb candidate "Fluggda" extracted from the sentence "Flag is fluttering among the towering masts". 25% sentence rate: "The flag was ragged at the end of the mast." In the case where the expression "stretched" includes 8%, the sentence proportion replaces "running" with the largest "floating".

In contrast, the translator 3-1 first translates "A black flag is flying high on its mast." To "A black flag is high on its mast." After extracting as a plurality of search nouns of the primary translation, search for it in the big data to collect sentences including the plurality of search nouns and classify each sentence as the first to n alternative candidate translations.

Subsequently, the translation unit 3-1 extracts a plurality of reference representative words from the primary translation, and measures the similarity between the plurality of reference representative words and the plurality of first to n alternative representative words and based on the result. By inferring the similarity between the verb and the verb replacement candidates of 1 to n, the verb substitution candidate having the largest similarity is replaced with the verb.

The measurement of the similarity between the verb and the plurality of first to n substitute candidates may be performed by checking whether a common word exists between the plurality of reference representative words and the plurality of first to n alternative representative words. It may be calculated according to Equation (1) below.

(Equation 1)

Here, n represents the number of substitute representative words extracted from one substitute candidate translation. At this time, the Si value has a value of 1 when the i-th word of the substitute representative word is commonly present in the reference representative word and the replacement representative word.

The Si value has a value of zero when the i th word of the reference representative word is not present in the replacement representative word.

Next, the concept image generator 3-2 directly generates a concept image according to translation.

Here, the concept image refers to an image produced so that the concept is sequentially reminded according to the translation.

In addition, the concept image generation unit 3-2 may additionally generate and provide a direct explanation symbol description image, a claim (negative) meaning unit display image, a conjunction verb display image, a negative phrase display image, and the like.

As an example, in Figure 4, if the translation is "flag / flap / on mast /", the flag image is first shown, then the flag is fluttered, and then the mast is In addition, it shows how the English sentences are organized in a conceptual image like this, so that the composition of the English sentences can be conceptually understood.

In addition, the concept image generation unit 3-2 may directly provide a symbol description image and additionally provide a claim (negative) meaning unit display image, a conjunction verb display image, a negative phrase display image, and the like.

In addition, the comparison image generator 3-3 generates a comparison image, which is designed to sequentially associate concepts according to the Korean word order translation.

As an example, referring to FIG. 6, the Korean word order translation shows a flag image first when the flag is fluttering on the mast, and then shows that the flag is on the mast, and then flutters. In addition, it shows the order in which the Korean sentences are constructed in a comparative video such that the difference between the structure of the English sentence and the structure of the Korean sentence can be understood.

7 is a view showing the configuration of a translation unit according to the present invention.

Referring to FIG. 7, a translation unit according to the present invention includes a sentence divider 10, a direct reading translator 12, a Korean word order translator 14, an word order sorter 16, and a post-processing performer 18.

First, the sentence divider 10 extracts all possible split point candidates by reading an input sentence from beginning to end. Here, the split point candidates used are punctuation marks, modifier phrases, and modifier clauses.

Here, punctuation marks include periods (.), Question marks (?), Exclamation points (!), Commas (,), suffixes (·), dashes (:), semicolons (;), dashes (/), and double quotation marks ( “”), Single quotes (''), parentheses (()), braces, ({}), brackets ([]), ellipses (-), ellipses (-), tildes (~), ellipses (..) ....).

At this time, the modifier or modifier clause is directly read by the symbol '/', '△' or '○'.

However, if the noun or noun clause is continued in the case of a comma, the sentence divider 10 excludes the split point candidate. In this case, the sentence divider 10 does not exclude a split point candidate when the word following the comma is a preposition.

In addition, the sentence divider 10 excludes the split point candidate when the word is the injection phrase in the case of the modifier phrase.

' 기호가 표시되어 있는 경우에 분할점 후보에서 제외한다.That is, in the case of the '/' symbol, the sentence divider 10 has a front word as a spray phrase '

If the symbol 'is displayed, it is excluded from the split point candidate.

In addition, the sentence divider 10 adds virtual split points before and after the verb to distinguish them.

That is, the sentence divider 10 adds an imaginary split point before and after the underline in the case of the verb.

The sentence divider 10 is excluded from the split point candidate when the verb is a next verb, that is, a noun.

However, the sentence divider 10 adds a virtual split point before and after the but in the iso conjunction.

That is, the sentence divider 10 is denoted by a '△' symbol when it is an equal conjunction, and is excluded from a candidate for split point when there is no underscore symbol representing a verb after the '△' symbol, but when it is but in an equivalent conjunction. If there are underscores, add virtual split points before and after.

Next, the direct translation translator 12 receives the sentence-splitting English sentence and translates the sentence in the order of sentence division.

Here, the direct reading translator 12 performs a semantic unit operation of the direct reading symbol in the English sentence with the direct reading symbol, and performs the direct reading on the semantic unit based on the reading.

The Korean word order translator 14 receives the direct reading word translated by the direct reading translator 12 and adjusts the order according to the Korean word order to generate the Korean word order translation.

At this time, the direct reading translator 12 translates a limited part defined by a relative pronoun or a relative adverb by a verb of the modified clause when the clause is a relative pronoun or a relative adverb.

'기호를 함께 표시되어 있는바, 직독직해 번역기(12)는 '○'기호에 '

'기호가 있는 경우에 한정 부분으로 수식어절의 동사를 대체하여 번역한다. That is, the sentence divider 10 displays the symbol '○' when the modifier clause is a related pronoun or a relative adverb and moves forward in the direction of a relative pronoun or a relative adverb.

'The symbol is displayed together, direct reading translator (12) is a' ○ 'symbol'

'If there is a symbol, translate it by replacing the verb of the qualifier clause with the limited part.

On the other hand, the Korean word order translator 14 may be reversed in the case of the modifier associated with the subject.

For example, if the English sentence is "On a clear day, a crewmember on a merchant ship sailing across the caribbean Sea peers out at the horizon through his telescope," on a clear day, a crew member flies his telescope on a merchant ship across the Caribbean. Stare at the horizon through the words.

To prevent this mistranslation, word order sorter 16 extracts nouns before the verb.

At this time, the word order sorter 16 extracts nouns with strong semantic elements, and the word order sorter 16 selects a crew member and a merchant ship as an example. As such, the word order sorter 16 performs the following operation when there are at least two nouns before the verb.

In addition, the word order sorter 16 inputs a crew member and a merchant ship from the big data, extracts sentences including the corresponding word from the big data, calculates a sentence ratio in which the merchant ship is represented after the crew member in the extracted sentences, and vice versa. Following the merchant ship, the crew calculates the sentence proportions expressed and prepares the Korean word order translations arranged in the order of the larger sentence proportions, and provides them to the user through the content transmission unit 4.

The word order sorter 16 has a 30% sentence rate in which the merchant ship is expressed after the crew in the sentences extracted as an example, and the sentence rate calculated by calculating the sentence rate in which the crew is expressed after the merchant ship is 70%. The word order aligner 16 produces a Korean word order translation translated, "On a clear day, a crew stares at the horizon through his telescope on a merchant ship crossing the Caribbean Sea."

Meanwhile, the post-processing executor 18 performs a verb post-processing process to correct translational errors with respect to the verb.

To this end, the post-processing performer 18 extracts the search nouns before and after the verb and searches for and extracts sentences including related search nouns from the big data.

The post-processing processor 18 extracts a verb expression having a meaning similar to that of the verb as a verb replacement candidate, calculates a sentence ratio with respect to the extracted verb replacement candidates, and then substitutes the largest verb having the largest sentence ratio. Substitute the verb expression as a candidate.

In one example, post-processing implementer 18 primarily translates "A black flag is flying high on its mast." To "A black flag is high on its mast." After extracting a plurality of representative nouns in the primary translation, the search is performed on the big data to collect sentences including the plurality of representative nouns.

At this time, the sentences that can be collected by the post-processing performer 18 are, for example, "the masts are torn apart by the wind torn flags. The flag flutters among the towering masts." At the end of the mast there was a terribly torn flag hanging. ”The post-processing executor 18 extracts a verb expression having a meaning similar to that of the verb as a verb replacement candidate to calculate a sentence ratio.

As a result of calculating the sentence ratio, the post-processing performer 18 includes, for example, "dancing sad dance", which is an alternative verb candidate extracted from "the masts torn apart in the wind. 2% sentence rate, including the alternative verb candidate "Fluda", extracted from the sentence "Flag is flung between the towering masts." If 8% of the expressions include "stretched" in the expression, the sentence proportion replaces "running" with the largest "floating".

In contrast, post-processing executor 18 primarily translates "A black flag is flying high on its mast." To "black flag is high on its mast." After extracting a plurality of search nouns of the primary translation, the search is performed on the big data, and the sentences including the plurality of search nouns are collected to classify the sentences as alternative candidate translations of the first to n.

Subsequently, the post-processing performer 18 extracts a plurality of reference representative words from the primary translation, and measures the similarity between the plurality of reference representative words and the plurality of first to n alternative representative words, based on the result. By inferring the similarity between the verb and the verb replacement candidates of 1 to n, the verb substitution candidate having the largest similarity is replaced with the verb.

The measurement of the similarity between the verb and the plurality of first to n substitute candidates may be performed by checking whether a common word exists between the plurality of reference representative words and the plurality of first to n alternative representative words. It may be calculated according to Equation (1) below.

(Equation 1)

Here, n represents the number of substitute representative words extracted from one substitute candidate translation. At this time, the Si value has a value of 1 when the i-th word of the substitute representative word is commonly present in the reference representative word and the substitute representative word.

The Si value has a value of zero when the i th word of the reference representative word is not present in the replacement representative word.

Features, structures, effects, and the like described in the above-described embodiments are included in at least one embodiment of the present invention, and are not necessarily limited to only one embodiment. Furthermore, the features, structures, effects, and the like illustrated in each embodiment are described. The embodiments may be combined or modified with respect to other embodiments by those skilled in the art.

Therefore, it should be interpreted that the contents related to such a combination and modification are included in the scope of the present invention. In addition, the above description has been made with reference to the embodiments, which are merely examples and are not intended to limit the present invention, and those skilled in the art to which the present invention pertains may be illustrated as above without departing from the essential characteristics of the present embodiment. It will be appreciated that various modifications and applications are possible. For example, each component shown in detail in the embodiments may be modified. And differences relating to such modifications and applications will have to be construed as being included in the scope of the invention defined in the appended claims.

1: Translation System
2: content input unit
3: direct symbol generator
3-1: Translation Department
3-2: Concept image generation unit (including direct symbol description image generation unit, claim (negative) meaning unit display image generation unit)
3-3: Comparative Image Generator
4: content transmission unit

Claims (15)

A content input unit for receiving English content from a user terminal;
A direct reading symbol generator for adding a direct reading symbol to the English content inputted from the content input unit;
A translation unit configured to receive an English sentence including a direct reading symbol generated by the direct reading symbol generator and perform a direct reading to generate a direct reading translation; And
It includes a content transmission unit for providing the English content and the direct-to-direct translation to the user terminal is added to the direct reading symbol generated from the direct reading symbol generation unit,
The translation unit receives an English sentence containing a direct reading symbol generated by the direct reading symbol generator, performs a Korean word order translation, and generates a Korean word order translation sentence.
The translation unit
A sentence divider for dividing a sentence by extracting all possible splitting point candidates while reading an input sentence from beginning to end;
A direct reading translator which receives the sentence-splitting English sentence in the sentence divider and translates it according to the order of sentence division to generate a direct reading translation;
A Korean word order translator which receives the direct translation word translated by the direct reading word translator and generates a Korean word order translation by adjusting the order according to Korean word order; And
And a post-processing performer performing a verb post-processing process for correcting a translation error in relation to a verb of a Korean word-order translation of the Korean word-order translator. delete The method according to claim 1,
Further comprising a concept image generation unit for generating a concept image produced to relate to the concept sequentially according to the direct translation generated by the translation unit,
The system of claim 1, wherein the content transmitter provides the concept image to a user terminal. The method according to claim 3,
The apparatus may further include a comparison image generation unit configured to generate a comparison image, the concept image being sequentially associated with the Korean word order translation sentence generated by the translation unit.
The system of claim 1, wherein the content transmitter is provided with a machine learning-based direct reading symbol for providing the comparison image to a user terminal. The method according to claim 1,
When the text input unit receives text, the English sentence is extracted and provided to the direct reading symbol generator. In the case of an image or a video, the content input unit extracts the English sentence included in the image or video. English sentence translation providing system with direct reading based on machine learning provided to the generation unit. The method according to claim 1,
The direct reading symbol generating unit analyzes the morpheme based on the inputted English content, displays a specific symbol for each part of speech, and generates a sentence processing sentence data by generating a symbol processing sentence data. . The method according to claim 6,
The direct reading symbol generating unit analyzes a sentence component consisting of two or more combinations of a subject, a descriptive word, a target word, a bore, and other formula words to determine a sentence form of any one of the first to fifth forms, and accordingly the sentence component and sentence Generates first result data in a form, and receives the first result data in which sentence components and sentence forms of sentence data are analyzed, such as nouns, verbs, modifiers, infinitives, nouns, participles, comparative grades, rank conjunctions, conjunctions, relative pronouns, Analyze sentence components using one or a combination of parts-of-speech groups including related adverbs to generate second result data that classifies sentence data by parts of speech, and deliver the first result data and the second result data. Obtain a symbol based on machine learning to generate symbol processing sentence data by displaying specific symbols for each part of speech. An English sentence translation system available. The method of claim 7, wherein
And a specific symbol generated by the direct reading symbol generator by underlining the corresponding word or phrase, or by using a direct learning symbol based on machine learning to display the original letter, triangle, or arrow. delete The method according to claim 1,
And the split point candidates are punctuation marks, modifier phrases and modal clauses. The method according to claim 1,
And a word order sorter for aligning a position of a modifier related to a subject in the Korean word order translation sentence. A content input unit for receiving English content from a user terminal;
A direct reading symbol generator for adding a direct reading symbol to the English content inputted from the content input unit;
A translation unit configured to receive an English sentence including a direct reading symbol generated by the direct reading symbol generator and perform a direct reading to generate a direct reading translation; And
It includes a content transmission unit for providing the English content and the direct-to-direct translation to the user terminal is added to the direct reading symbol generated from the direct reading symbol generation unit,
The translation unit receives an English sentence containing a direct reading symbol generated by the direct reading symbol generator, performs a Korean word order translation, and generates a Korean word order translation sentence.
The translation unit
A sentence divider for dividing a sentence by extracting all possible splitting point candidates while reading an input sentence from beginning to end;
A direct reading translator which receives the sentence-splitting English sentence in the sentence divider and translates it according to the order of sentence division to generate a direct reading translation;
A Korean word order translator which receives the direct translation word translated by the direct reading word translator and generates a Korean word order translation by adjusting the order according to Korean word order; And
A word order sorter for sorting a position of a modifier related to a subject in the Korean word order translation,
The word order sorter extracts nouns in front of the verb, extracts sentences containing the word from the big data, calculates a sentence ratio for each noun order from the extracted sentences, and creates a Korean word order translation sentence arranged according to the calculated sentence ratio. English sentence translation providing system having a machine learning-based direct reading symbols provided to the user through the content transmission unit. delete The method according to claim 1,
The post-processing performer extracts a plurality of search nouns, retrieves and extracts sentences including related search nouns from the big data, extracts verb expressions having a meaning similar to that of the corresponding verbs as verb replacement candidates, and extracted verb replacement candidates. A system for providing English sentence translation based on a direct reading symbol that substitutes a verb expression after calculating a sentence ratio for the largest verb replacement candidate with the largest sentence ratio. The method according to claim 1,
The post-processing performer extracts a plurality of search nouns of the first translation sentence, searches the big data, collects sentences including the plurality of search nouns, and classifies each sentence as a first to n alternative candidate translation sentence. Extract a plurality of reference representative words from the primary translation, and measure the similarity between the plurality of reference representative words and the plurality of first to n alternative representative words, and replace the verbs with the verbs based on the result. A system for providing English sentence translation with direct reading symbols that infers similarity between candidates and replaces the verb substitution candidate with the highest similarity with the corresponding verb.

Images