KR20040007741A - Cross-idea association database creation - Google Patents

Abstract

A method and apparatus for generating a cross-ideal correlation database are disclosed. The cross-ideal database correlates words and phrases in one language corresponding to information in a state with words and phrases in a second language corresponding to information in a second state. The method includes receiving content represented in the first state and receiving content represented in the second state, and analyzing the content in the first state and the content in the second state. The analysis uses the content segments represented in the first state and the content segments in the second state. The method includes creating a correlation database of content in the first state relative to the content in the second state.

Description

Cross-ideal correlation database method and system {CROSS-IDEA ASSOCIATION DATABASE CREATION}

In one embodiment of the invention, two states represent a language (eg, English, Hebrew, Chinese, etc.) and the present invention provides a cross-language database that correlates words and phrases of a first language to translation counterparts of a second word. To generate. In this example, the present invention creates a database by examining documents in two languages and by creating a translation database for each word or phrase in both languages. Due to the present invention, the user creates a database of ideas and correlates these ideas to other ideas in a vertical fashion. Thus, ideas are correlated with other ideas and ranked according to their frequency of occurrence. The specific weights given to the frequency of occurrence, and the uses applied to the generated databases, may vary according to user requirements.

For example, in the conversion from one language to another, the present invention will operate to generate foreign language words and phrase translations for English words. The present invention will export a correlation ranking for the words (or phrases) when a sufficiently large sample size is provided. For example, the most frequently occurring words will be foreign language words equivalent to English words. However, the present invention will also export other language correlations with English words, and the user can manipulate the correlations as desired. For example, the word "mountain", when operating in accordance with the invention, may export a list of foreign language words of the language being examined. The foreign language corresponding to the word "mountain" will be ranked highest. However, the present invention will export other foreign language words related to "mountain", such as "snow" or "ski". These words, which may (or may not) be ranked lower than the translation of "mountain", may be manipulated as desired by the user. Thus, the present invention is an autocorrelation database generator. The largest correlation means "translation" in some respects, while the other frequent correlations represent ideas that are closely correlated to the idea under examination.

The present invention relates to a method and apparatus for creating a cross-ideal correlation database for transforming and manipulating information from one state to another.

It is an object of the present invention to develop a database of correlation of words and phrases between one language and a second language. In general, the method includes performing inspections and operations on two documents, where each document has text in two languages that represent the same concept or content. The method and apparatus according to the invention are used such that a database is created by the correlation between the two languages. Translations of words and phrases are basic and are used to create a database with correlations. Translations and other correlations for words and phrases between the two languages become stronger and more frequent as more documents are examined by the present invention. Thus, by operating on sufficiently rich document "samples", the most common (and most accurate) correlation will appear, allowing the method and apparatus to be used for translation purposes.

Preferred embodiments of the invention utilize computing devices such as personal computer systems of the kind known in the art. However, the method and apparatus of the present invention do not need to use such a computing device, and can be easily accomplished by other means, including manual generation of cross-correlation. The manner in which a series of documents are examined to magnify document "samples" and create a cross-correlation database varies. That is, documents can be set up for analysis and manipulation either manually or by automatic feeding (eg, an automatic paper feeder as known in the art) or by using a search technique on the Internet for automatic retrieval of a correlated document. Can be.

In the following description the term "documents" is used interchangeably to mean item pairs (eg, books, articles, spellings, etc.) representing the same concept. However, one in the pair is one language and the other is a second language. In addition, whenever the present invention is deemed to operate on words, it is obvious that the same techniques apply to phrases and are not limited to just one word.

This application cites US Provisional Application No. 60 / 276,107, "Method and Apparatus for Content Manipulation," dated March 16, 2001.

Preferred embodiments of the invention will now be described.

The invention can be used in a typical computer system including one or more display means, input methods and output methods, and a processor. The display means can be any of those available in the known art, such as cathode ray tube terminals, LCDs, flat panel display devices and the like. The processor means may use any of the available ones used in the computing environment, such that the processor means is supplied to operate the computer for the implementation of the present invention. Finally, the input method is used for inputting a document for the purpose of building a cross-correlation database, and as described above, the specific input method may vary according to user needs.

According to the invention, documents are checked for database construction. After inputting a document (of a pair of documents representing the same text in two languages), the generation process begins using the method and apparatus introduced herein.

To aid understanding, the documents contain the same word (ideology in comprehensive terms) in two languages. Document A is written in language A, and document B is written in language B. These documents have the following text:

Document A (Language A): X Y Z X W V Y Z X Z

Document B (Language B): AA BB CC AA EE FF GG CC

The first step of the present invention is to compute a word range that determines an approximate correlation for a given word or phrase. Because word-to-word translation is not appropriate (ie, it is likely that word 1 of document A does not exist as a literal translation of word 1 of document B), the database generation technique of the present invention provides Test words against a range of words in a second language. Thus, this range is developed by examining two documents and is used to compare words, phrases, or other word strings in the second document with words, phrases, or other word strings in the first document. That is, the word range of the second document is applied to match as much as possible with any word (or phrase, or word string) of the first document. By running tests for one range, the database generation technique builds a number of second languages that may be comparable translations for the first language words.

This range value is ultimately defined by the user. Various techniques can be used to determine range values, including common statistical techniques such as the declination of the bell curve based on the number of words in the document. Using statistical techniques such as the bell curve, the range at the beginning and end of the document will be smaller than the range in the middle of the document. The bell-shape frequency for range words enables rational extrapolation of possible word conversions, whether derived from a comprehensive percentage of the number of words desired or from the number of words in the document. Like a "stair" technique that allows a range to exist at a third level, such as a first level for a word in a percentage, a second level for a word in another percentage, and a first level for a word in the last percentage. There are other techniques for calculating ranges. In addition, the range is defined by the user according to other possible variables.

The range value can depend on the number of words in both documents. Any value can be given if the two documents have the same number of words. When applying statistical techniques, a bell curve can be generated such that the range is composed of a small number of words at the beginning of the document, the largest number of words in the middle of the document, and a small number of words at the end of the document.

If the words in the two documents are not the same, the ratio can be used to position this range correctly. For example, if document A has 75 words and document B has 100 words, the ratio between the two documents is 3: 4. The midpoint of document A is word position 37 (or 38). However, using this midpoint (word position 37 or 38) as the position for the largest range value of document B (if determined using the bell curve) is not effective. This is because this position (word position 37 or 38) is not the midpoint of document B. Instead, the maximum application point of the range value of Document B can be determined by the ratio of words between two documents, by manual positioning of the midpoint of Document B, or by some other technique.

The gist of the present invention is to generate a correlation frequency for each possible transform. By looking at the position of a word in the document, and by applying a range as described above, the database technology of the present invention exports the set of possible characters of the second language document to be converted into the words of the first document. As the database generation technique of the present invention is used, the possible character set will be narrowed, and the correlation frequency will be developed to help the potential conversion decision. Thus, after examining a pair of documents, the present invention will generate a correlation frequency of a word (or phrase, or word string) of one language relative to the same word (or phrase, or word string) of the second language. If multiple document pairs are examined in accordance with the present invention, so that many samples are generated, the cross-language correlation database generation technique will yield increasingly higher correlation frequencies for any single word, phrase, or word string. After enough samples have been reviewed, the highest correlation frequency appears in translation. Of course, the ultimate point that the correlation frequency is considered to be an accurate translation is defined by the user, such as other translations such as those introduced in US Provisional Application No. 60 / 276,107, "Method and Apparatus for Content Manipulation," dated March 16, 2001. Depends on the conversion technique. For example, after searching a large number of documents, the following correlation frequency can be obtained in Spanish equivalent to "friend": "gato"-25%, "burro"-15%, "amigo"-60% . As more pairs are examined, the operation of the present invention will increase the correlation frequency for "amigo" and decrease the frequency for "gato" and "burro". At the point defined by the user, a correlation frequency will be reached at which level the translation of the English word "freind" is considered to have been translated into Spanish "amigo".

As mentioned above, the invention tests phrases as well as words, as well as word strings. After one word has been analyzed, the database generation technique according to the present invention analyzes the two-word word string, and then the three-word word string, and so on, in an incremental manner. This technique allows you to translate a phrase or word string from one language to one word in another language. If the number of words is greater than 1, the analysis ends when all positions for the word or word string have been analyzed. If a word appears only once in the document, this process will immediately increment the word and export the word string. If the word string appears only once, this process will return back to the second word of the document, where the analysis cycle reappears as described above.

Changes to this process can be made to accommodate the case where a word appears only once in both documents to be examined. For example, if a word appears only once in a document, a change in the present invention allows analysis to be made on other documents to search for correlated words or word strings. In one aspect, any number of documents can be aggregated and treated as a single document for the operation of the present invention. In yet another embodiment, work can be done on the entire document to accommodate situations where words appear only once.

The same goes for word strings: increment, test, and return. Thus, the frequency count for a phrase is checked, the phrases are returned based on the range, and a possible translation database for that phrase is created.

In addition, the present invention may operate in such a manner as to analyze a word string that depends on exact positioning or words, and may operate in such a manner as to consider grammatical specificity such as a phrase, a style, or an abbreviation.

The present invention can accommodate many other changes that occur in documents where word subsets appear within larger word strings. For example, the exact name may be given completely (e.g. "John Doe"), may be called only by last name or first name (e.g. "Doe" or "John"), or may be abbreviated in another way (e.g. , "Mr. Doe"). The present invention is responsible for recognizing the presence of these patterns in the correlation database through analysis, and by manipulating frequency returns. The present invention is easier to export individual word returns than a word string return (i.e., the surname or first name only returns more than the full name "John Doe" word string), so that the words that make up the word string are phrases. As will be counted separately, as in part of, a change in rank may be used. For example, in any document, the name "John Doe" may appear a hundred times, "John" may appear 120 times, and "Doe" may appear one hundred times. A normal translation return will rank "John" higher than "Doe", and both words will rank higher than the word string "John Doe" (both when attempting to parse a single string called "John Doe"). . By subtracting the largest word string frequency from the return for the subset, proper ordering can be achieved (other methods can of course be used to achieve similar results). Thus, subtracting 100 from 120, the corrected return for "John" is 20. Applying this analysis, one can get 100 as the frequency for "John Doe", 20 as the frequency for "John", and 10 as the frequency for "Doe", thus generating an appropriate correlation.

One embodiment of the invention will now be described using the two documents described above. The table is regenerated as follows.

Document A (Language A): X Y Z X W V Y Z X Z

Document B (Language B): AA BB CC AA EE FF GG CC

Using the two documents listed above, the following steps occur for a database creation technique.

Step 1: First, the range is determined. As mentioned above, the range can be approximated by various methods or defined by the user. The word counts for the two documents are approximately the same (10 words in document A and 8 words in document B). A range of 3 (30% of the words in document A) can provide the best results. In this example, to approximate the bell curve, the range will be one at the beginning and end of the document, and two at the middle of the document. However, as mentioned above, the range may be defined by the user as a whole.

Thus, in the case of this example, as the database technology of the present invention is used, the range may change from one word to two words, to one word.

Step 2: Then, the first word of document A can be examined and tested for document A to determine the frequency of the word in the document. In this example, the first word of document A is X. X appears three times in document A (positions 1, 4, 9). The position number of a word, phrase, or word string is a simple indication of the number of times a word, phrase, word string is presented in the document and the position of the word, phrase, or word string in the document relative to other words. Thus, the location number corresponds to the number of words in the document and ignores punctuation. For example, if there are 10 words in the document, and the word "kick" appears twice, the position number of the word "king" is simply the position (of 10 words) where the word appears.

Since the word X appears more than once in the document, the process goes to the next step. If the word X appears only once, the word is skipped and the process proceeds to the next word string (or phrase) phase and the generation process continues.

Step 3: A possible second language translation for the first language word X in position 1 is returned. Applying this range to Document B indicates the words AA and BB at position 1 and position 2 (1 +/− 1) of document B. All possible combinations of this word revert to potential translations for X: AA, BB, and AA BB (word string combinations). The word string combination returns to a possible match to accommodate the fact that one word of the w 1 language can be equivalent to a phrase of the second language. Thus, X1 (first occurrence of word X) exports AA, BB, AA BB as correlators.

Step 4: The next position of the word X is analyzed. This word X2 appears at position 4. Because position 4 is in the center of the document, the range will be two words. Possible translations are shown by looking at word 4 of document B and applying its range (2). Thus, two words before word 4 and two words after word 4 appear. Thus, words at position 4 +/- 2 appear, or at position 2, 3, 4, 5, 6. These positions correspond to the words BB, CC, AA, EE, and FF in document B. All forward permutations of these words are now considered. That is, BB, CC, AA, EE, FF, BB CC, BB CC AA, BB CC AA EE, BB CC AA EE FF, CC AA, CC AA EE, CC AA EE FF, AA EE, AA EE FF, EE FF is considered. Thus, X2 is BB, CC, AA, EE, FF, BB CC, BB CC AA, BB CC AA EE, BB CC AA EE FF, CC AA, CC AA EE, CC AA EE FF, AA EE, AA EE FF, Export EE FF as a correlation.

Step 5: The return of the first occurrence of X (position 1) is compared with the return of the first occurrence of X (position 4) and a match is determined. In this case the correlations for X1 and X2 are compared and a match of the two documents is provided. Identical returns (or word occurrences or word strings) in the overlap region between two ranges can be reduced to a single occurrence. For example, in this example where the word at position 2 is BB, it appears the same for the first occurrence of X and the second occurrence of X. Since this same word position appears for both X1 and X2, words are counted in one occurrence. However, even if the same word appears, if it is not placed within the overlapping zone (ie, the same word position does not appear for X1 and X2 but the result appears with the same word), the word is considered twice. In this case the return for word X is AA. This is because the word AA occurs in both the correlation returns for X1 and X2. The other word that occurs in both correlation returns is BB. However, as mentioned above, since the word is in the same position reached by the range operation for the first and second occurrences of X, this word can be ignored.

Step 6: The next position of the word X (position 9) X3 is analyzed. Applying a range of 1 (near the end of the document) results in the values of the following positions in document B: 8, 9, 10. Since document B has only eight positions, the result is truncated and only word position 8 appears as a possible value for X. That is, CC appears.

There is no match when compared to the first return for X (X1). Thus, since no match is shown, the value shown for X3, i.e., CC, is ignored and no correlation match is provided.

Step 7: The next position for word X is analyzed. However, in document A the word X no longer appears. At this point a correlation frequency is built for word X, and the following database is created with a possible translation (AA) for X. Thus, at this point, there is a correlation of X to AA.

Step 8: Since there is no further occurrence of word X, the process is incremented word by word and the word string is tested. In this case the word string being checked is "X Y", which is the first two words of document A. The same technique described in steps 2-7 applies to this phrase.

Step 9: By looking at document A, we can see that the word string X Y occurs only once. At this point, the incremental process stops and no database creation takes place. Since the end point is reached, the next word is checked (this happens whenever there is no match for the word string). In this case, the word at position 2 of document A is "Y".

Step 10: Applying the steps 2-7 to the word "Y", the following results are obtained.

The word Y appears twice (positions 2 and 7), and so the database creation process continues (and if Y appears once in document A, Y is not checked).

The range of position 2 is 1 word.

Applying a range to document B (position 2, the first occurrence of word Y) results in positions 1, 2 and 3 of document B.

-The corresponding foreign language words in this position are AA, BB, CC.

Applying forward permutation leads to the following possibilities for Y1. That is, AA, BB, CC, AA BB, AA BB CC, BB CC.

The next position of Y is analyzed (position 7).

The range of position 7 is two words.

Applying this range to document B (position 7) results in positions 5, 6, 7, and 8. EE FF GG CC.

All permutations yield the following possibilities for Y2: EE, FF, GG, CC, EE FF, EE FF GG, EE FF GG CC, FF GG, FF GG CC, GG CC.

The match result from Y1 exports the CC as a single match.

Combining the correspondences for Y1 and Y2 yields a CC with a correlation frequency for Y at one value.

Step 11: Because the only possible match for word Y (word CC) occurs at the end of the range for the first occurrence of Y (CC occurring at position 3 of document B), the range is position 1, 2, To export 3, 4, or the following forward permutations: AA, BB, CC, AA BB, AA BB CC, AA BB CC AA, BB CC, BB CC AA, CC AA, to export 1 by 1 Incremented. Applying this result, one can still derive CC as a possible translation of Y. The range was incremented because the indicated matches were at the end of the range for the first occurrence. Each time this pattern occurs, the end of the range increment will occur as a substep (or alternative step) to ensure completion.

Step 12: Since there is no increment of "Y" in document A, the analysis increments one word of document A and the wordstring "Y Z" is checked (word after word Y). Incrementing to the next string (Y Z) and repeating the process yields:

In document A, the word string Y Z appears twice: position 2, position 7.

The possibility for Y Z in the first occurrence (Y Z1) is AA, BB, CC, AA BB, AA BB CC, BB CC.

The possibility for Y Z in the second occurrence (Y Z2) is EE, FF, GG, CC, EE FF, EE FF GG, EE FF GG CC, FF GG, FF GG CC, and GG CC.

Matches and combinations derive CC as a possible translation for the word string Y Z.

Range extension (end of range increment) yields the following results for Y Z: AA, BB, CC, AA BB, AA BB CC, BB CC, BB CC AA, CC AA.

Applying this result, we derive CC as the correlation frequency for the word string Y Z.

Step 13: Since there is no further occurrence of "Y Z" in document A, the analysis increments one word of document A and the word string "X Y Z" is checked (word after word Z in position 3 of document A). Repeating this process incrementing to the next phrase (Y Z X) (two occurrences of Y Z X in document A) yields the following result:

-Range is 2. Because the midpoint of the phrase appears closer to the midpoint of the document.

The result of the first occurrence of Y Z X is in positions 2, 3, 4 and 5.

The permutations are BB, CC, AA, EE, BB CC, BB CC AA, BB CC AA EE, CC AA CC AA EE, and AA EE.

The result of the second occurrence of Y Z X is at positions 5, 6, 7, 8.

The permutations are EE, FF, GG, CC, EE FF, EE FF GG, EE FF GG CC, FF GG, FF GG CC, GG CC.

Step 14: Increment to the next word string (Y Z X A) finds only one occurrence. Thus, the word string database generation is completed and the next word Z is checked (position 3 of document A).

Step 15: Applying the above steps for Z appearing three times in Document A, the following results are obtained.

-Returns to Z1 are AA, BB, CC, AA, EE, AA BB, AA BB CC, AA BB CC AA, AABB CC AA EE, BB CC, BB CC AA, BB CC AA EE, CC AA, CC AA EE, AA EE.

The return for Z2 is FF, GG, CC, FF GG, FF GG CC, and GG CC.

Comparing Z1 and Z2 leads to CC as possible matches.

CC is derived by the correlation frequency for the word Z by returning Z3 and comparing with Z1.

Step 16: The increment to the next word string results in the word string Z X, which appears twice in document A. Applying the steps described above for Z X yields the following results.

Returns for Z X1 are BB, CC, AA, EE, FF, BB CC, BB CC AA, BB CC AA EE, BB CC AA EE FF, CC AA, CC AA EE, CC AA EE FF, AA EE, AA EE FF, EE FF.

The returns for Z X2 are FF, GG, FF GG, FF GG CC, GG CC.

-Through the comparison process, the correlation of the word string Z X is represented as CC.

The comparison with the return for Z X yields the CC as the correlation frequency for the forward string Z X.

Step 17: The next phrase incremented is Z X A, which occurs only once, so that the next word X of Document A is checked.

Step 18: The word X has already been checked in the first position. However, compared to other documents, the second position of word X was not checked for possible returns for word X. Thus, word X (in the second position) behaves as in the first occurrence of word X and proceeds forward in the document.

The return for X in position 4 is BB, CC, EE, FF, BB CC, BB CC AA, BB CC AAEE, BB CC AA EE FF, CC AA, CC AA EE, CC AA EE FF, AA EE, AA EE FF and EE FF.

The return for X in position 9 is GG, CC, GG CC.

Comparing the result of position 9 to the result of position 4 yields a CC with a possible match for word X.

Comparisons with the returns for X yield a CC with a correlation frequency for word X.

Step 19: Dividing by the next word string, the word string X A is obtained. However, this word string does not appear more than once in document A. Thus, the process examines the next word (A). The word "A" appears only once in document A, and therefore an increment occurs. The increment does not proceed to the next word string. This is because the word "A" appears only once, and thus goes on to the next word "B" of document A. The word "B" appears only once in document A, and therefore the next word Y is checked. Since the word "Y" does not appear at a position higher than position 7 of document A, the next word (Z) is negated. The word "Z" appears twice in document A: position 8 and position 10.

Step 20: Applying the above procedure to the second occurrence of the word Z, the following results are obtained.

The return for Z in position 8 gets FF, GG, CC, FF GG, FF GG CC, GG CC.

Return to Z at position 10 to get CC.

Comparing position 10 to position 8 yields no match for the word Z.

In addition, the word CC appears as a possible match. However, since CC represents the same word position reached by analyzing the Z at position 8 and the Z at position 10, the match is ignored.

Step 21: Increment by one word yields the word string Z X. This word string does not appear anywhere else in Document A. Thus, the process starts again at the next word "X" in document A. The word "X" does not appear anywhere in Document A, so the process starts over again. However, the analysis ends when the end of document A is reached.

Step 22: The final correlation analysis is tabulated combining all the results from the above.

Insufficient data to export results for other words and phrases in document A. Individual possible words or word strings in document A result in several possible correlations for word CC in document B. As more document pairs with the word CC of language B are examined, the correlation frequency will increase the statistical confidence, so there will be one word (or word string) in translation for the word CC.

In another embodiment, the database generation techniques of the present invention can be used in a variety of ways to generate cross-language correlation. For example, a database can be created by simply pairing all words and word strings that occur in Document A with a range of words in Document B, without comparing them to multiple occurrences of words, and without using range increment techniques. have. This method uses the cross-language correlation principle to create a database in a different way than described above.

As an example of this embodiment, consider an example of two documents displaying the same concept or content in different languages.

Document A: The sky is blue. The grass is green. The sky includes clouds and stars. The grass dies in the winter.

Document B: AAA BB CCC. AAA EEE DDDD. AAA BB FFF GGGG HHH. AAA EEE III JJJJJ.

As a first step in this embodiment, the word count of each document can be constructed to generate the appropriate ratio. This ratio is used for comparative range positioning as introduced below. In this example, document A contains 20 words, document BB contains 15 words, and the ratio is 4: 3. Thus, the four words of document A correspond to the three words of document B.

As a next step, a word segment is constructed for the word string or phrase to be examined. This segment may be determined according to common language rules. For example, a segment may be a sentence or a paragraph. However, the length of the segment is defined by the user, and any portion of the desired word strings is possible. In the case of this example, the segments correspond to the sentences of each document, but longer segments than the single sentence are much more effective for generating the correlation of the present invention. Because the base of the potential correlation to populate the database is much larger.

As a next step, let's examine the first word of the first segment. In this example, the first word of the first segment is "the".

As a next step, all occurrence positions of the first word of document A are determined. The word position is determined by each word count position in any document. Using this example, the spaces of the word "the" are 1, 5, 9, 15.

As a next step, determine target words correlated to the first word being examined. Target words are determined by using word ratios to determine the relative points of document B, and by applying a range to the word position of document B. The relative position of a word in document B is determined by applying the above mentioned ratio. In this example, the word "the" appears at positions 1, 5, 9, and 15 of document A. These positions correspond to relative positions 1, 4, 7, 11 of document B. This calculation is made by taking the position of document A, building the ratio (by multiplying the word ratio of document B to document A or by multiplying 3/4), and applying the ratio. 1 (Document A) x 3/4 = 1 (rounded up); 5 (document A) x 3/4 = 3 3/4 = 4 (rounded up) (document B); 9 (document A) x 3/4 = 7 (document B, rounded up); 15 (document a) x 3/4 = 11 (document B, down).

When applying the above to the first word (the) to be examined, the following results can be obtained.

-Position of document A = 1

-Relative position of document B = 1

The frequency range applied to the front / back words of document B corresponds to word positions 1-3 of document B. This determination is made by taking the position +/- frequency range, or 1 +/- 2, or -1 to 3. By ignoring negative and zero positions, the word position results in document B appear as 1-3.

Applying this frequency range to document B results in words at positions 1, 2 and 3 of document B. That is, AAA, BB and CCC are shown.

Thus, the first occurrence of the word "the" in document A derives the words AAA, BB, CCC of document B.

As a next step, proceeding to the next occurrence of the word "the" in document A, and applying the preceding procedure,

Position of Document A = 4

Relative Position of Document B = 3

The frequency range (+/- 2) at relative position 3 leads to positions 1, 2, 3, 4 and 5 of document B. That is, AAA, BB, CCC, AAA, and EEE are derived.

It is then determined whether the target word for the second location matches the target word for the first location.

The result from the first search is AAA BB CCC.

The result from the second search is AAA BB CCC AAA EEE.

-Matches are AAA (twice), BB, CCC.

These matches are stored in the memory element for possible correlation with the word "the".

The process is repeated for the next occurrence of the word "the". The result from the third occurrence of the word "the" is CCC AAA EEE DDDD AAA. The match terms are AAA (twice) and CCC. The match is stored in a memory element for possible correlation.

This process is repeated for all other occurrences of the word "the". The results from this analysis show AAA BB FFF GGGG HHH and AAA and BB in a possible correlation.

As a next step, the present invention increments the number of words being checked one by one. In this example, the word being checked is "the" (the first word of document A). The next word string to be incrementally analyzed is "the sky".

The above steps are repeated for this word string. "The sky" appears in positions 1 and 9. The relative positions in document A are 1 and 6. Applying this frequency range to the relative location number, AAA, BB, CCC, and AAA EEE DDDD AAA BB are derived for the first location. Comparing the two results for two word phrases yields AAA and BB as possible correlations to be stored in the database.

The process is incremented by one word, so the process repeats for "the sky is". This process derives only the AAA BB CCC as a possible correspondence. This is because there is no other occurrence.

Repeating the process for the phrase "the sky is blue", there is only one occurrence with AAA BB CCC as a possible correlation to be stored in the database.

As a next step, we have reached the end of the first segment defined by the user, as indicated by the punctuation in Document A. The next step is to continue the iterative process described above by taking the second word of the first segment. In this example, the analysis includes "sky", "sky is", "sky is blue", resulting in 2 and 7 relative positions in document B, and AAA BB CCC AAA as the first match, The second match yields the EEE DDDD AAABB FFF. AAA and BB are derived as possible correlations to be stored in the database.

The next incremental words in the segment are "is" and "is blue". Repeating the above process for "is" and "is blue" yields AAA BB CCC AAA and CCC AAA EEE DDDD AAA as a match, where AAA and CCC are possible correlations to be stored in the database.

The next incremental word in the segment is "blue", which yields AAA BB CCC AAA EEE as a possible correlation to be stored in the database.

The analysis reached the end of the segment. The next segment is "the grass is green." Since "the" has already been analyzed, the next part of the word to be analyzed is "the grass". Then, "the grass is", "the grass is green", "grass", "grass is", "grass is green" and "green".

The process continues with the next segment ("the sky includes clouds and stars"), and analysis continues with "the sky includes clouds", "the sky includes clouds", "the sky includes and", "the sky includes clouds and stars", " About includes "," includes clouds "," includes clouds and "," includes clouds and stars "," clouds "," clouds and "," clouds and stars "," and "," and stars ", and" stars " Is done.

Finally, the process continues to the next segment ("the grass dies in the winter"), and analysis continues with "the grass dies", "the grass dies in", "the grass dies in the", "the grass dies in the winter "," grass dies "," grass dies in "," grass dies in the "," grass dies in the winter "," dies "," dies in "," dies in the "," dies in the winter " , "in", "in the", "in the winter", "the winter", and "winter".

In view of the foregoing that the segments need not be limited to sentences or paragraphs, the segments can be extended for analysis. When the present invention contributes to the function of language translation, several sentences can be analyzed into one segment by extending the segment to incorporate the person into the first sentence ("Went to school today. She walked to the school on the street ").

As mentioned above, these two embodiments represent techniques used to generate the correlation. The techniques of the present invention need not be limited to language translation. In a broad sense, these techniques will apply to any two embodiments of the same idea that can be correlated. This is because foreign language translation exists as a couple of ideas (words or phrases). Thus, the present invention encompasses not only data, sound, music, and video, but also a broad range of other concepts that exist as one idea, including ideas that can represent any sensory experience (eg, sound, sight, smell, etc.). Can be applied to correlate. All that is required is that the present invention analyze two examples. In language translation, the embodiments are documents, and in the case of music, the embodiments may be digital representations of sound waves and musical scores representing the same song.

In addition, one embodiment of the present invention may be loaded with any correlation into the database by mechanical, electrical, or other means. For example, loading a database with the foreign language counterparts of the English words it, his, her, an, a, of, or other common words to create a correlation database more accurately, efficiently, and with faster resolution. It is possible. Therefore, within the present embodiment, the present invention can automatically export foreign language counterparts of certain words loaded into the database. This embodiment can accommodate common words that may deviate from the analysis using the correlation database generation technique of the present invention.

Additionally, one embodiment may use a common correlation to generate and recognize word patterns. For example, like "President" for "Clinton", it is possible to load a correlation into the database so that the correlation database accommodates situations where the text means President Clinton but the word "president" is used as an abbreviation.

When cross-language correlation exists in a broad sense, such as a cross-ideal correlation technique for generating a possible correlation database, the result can be manipulated when the correlation is established. Thus, for example, when each "idea" is assigned a correlation to electromagnetic waves, it will be possible to create an "electromagnetic association" of the idea. When a given number of ideas have been encoded with the corresponding electromagnetic correlation, the data can be manipulated with electromagnetic waves and transmitted over existing network structures. When the electromagnetic wave reaches the receiving device, the receiving device will synthesize the waves into separate components and, under the correlation, will present the individual ideas represented by the electromagnetic correlation.

Claims (2)

In the content correlation method, Receive content expressed in a first state, Receive content expressed in a second state, Analyze the content represented by the first state together with the content represented by the second state, wherein the analysis uses content segments of the first state and content segments of the second state, and Create a correlation database of the content of the first state with respect to the content of the second state Content correlation method comprising the above steps. In a computer system for content correlation, the system An arithmetic device that receives the content represented in the first state and the content represented in the second state Wherein the computing device analyzes the content represented by the first state together with the content represented by the second state by using content segments of the first state and content segments of the second state. , And And said computing device generates a correlation database of said content in said first state relative to said content in said second state.