TW589549B - Method for calculating conceptual similarity based on common sense tree marking method - Google Patents

Abstract

A method for calculating conceptual similarity based on common sense tree marking method comprises the following procedures: firstly provide a common sense tree structure, which provides multiple common sense trees, subsequently provide a first glossary and second glossary, in which the first glossary provides at least a first definition and the second glossary provides at least a second definition, then determine if the first definition and the second definition are located at the same common sense tree, when the first definition and the second definition correspond to the same common sense tree, then determine at least a score value of the corresponding common sense tree, finally determine the similarity of the first glossary and the second glossary based on the score value.

Description

[Technical field to which the invention belongs] Particularly rare! Γ Establishing a Thesaurus is related to a method based on common sense tree marks and antonyms. [Principal technique] and antonym thesaurus methods, notations that calculate conceptual similarity and that are not able to express liguria similarity, meaning words, value thesaurus [thesaurus in the invention can usually only indicate that two Defined similarity, such as file similarity comparison and file efficiency services. For example, the old two include "person π, ', teaching", and can not quickly and objectively add vocabulary. Synonyms, so search engines and automatic summaries will not be able to provide teachers and professors although they are not the same as the definition of education ", resulting in rules. Some similarities are implemented by common sense tree. It includes the knowledge tree. First, the present invention establishes the main thesaurus of common sense based on the method of the invention and the above purpose. The concept similarity tree marking method t calculates the cost. The root of the standard example The following steps first provide a common sense and then provide a definition of the second word definition and the second definition of the score value similarity. The vocabulary and the second vocabulary have at least a second definition that is the same. Finally, the antonym can be used to calculate the approximate tree-like knot vocabulary.-The second is located in the common sense tree. In order to achieve a similar structure, the first word is defined. One basis. According to the method of the present invention, there are plural constants with at least the next common sense tree, when it is decided to correspond to the first vocabulary and judge the above __ the first definition and the at least the cognition tree-the second vocabulary

589549 V. Description of the invention (2) In order to make the above and other objects, features, and advantages of the present invention more comprehensible, the following describes the preferred embodiments and the accompanying drawings in detail, as follows: [ Embodiment] Please refer to FIG. 1. FIG. 1 is a schematic diagram showing an operation flowchart of a method for calculating conceptual similarity according to a common sense tree marking method according to an embodiment of the present invention. ‘First’, as step S1 〇, extract two words Take two = sinks in one article, such as teachers and mentors, a common sense tree structure with a plurality of common sense trees, such as an entity common sense tree, an 11 ribute common sense tree, and an event ( even) the common sense tree and a second feature (secondary, please refer to the attachment), the common sense tree is a subset of the a used in computer systems, contains definitions of different words, one of which has higher similarity at the lower level in a common sense tree; common sense Tree structure query contains the two word discards; according to-find in the knowledge tree-definition, | · bit 'event common sense tree; f two fixed = constant, the second feature common sense tree is found "Yiyi teacher "In the common sense tree structure, p A1 Dan, Month, and In addition, two definitions are found in the vocabulary" Guide ^ tta, J ^^^ ^ i, "and" teach ancient n. "The 2nd characteristic common sense tree teacher's office Included definitions: ^ ° humanl person-entity common sense tree

589549 V. Description of the invention (3) occupat ion | Position—Signal often teach | Teaching-event common sense tree " education | Education ~ 2nd tutor includes definitions: & human | person-entity common sense tree guide | guide-event Common sense tree desired | Good-Second feature common sense tree teach | Teaching-Event common sense tree 5 education | Education_Second feature common sense tree Next, step S14, judging that the two words contain the same common sense tree, and place the two words in The included definitions are driven by me—whether they are located on the same common sense tree and whether the two definitions are located on the same common sense tree. Steps are compared to determine whether they are located on the same common sense tree. For literally the same definition and the same meaning, I "will determine that the two definitions are the same. A score will be given to the two definitions of the regular bound tree. = The text spelling method means that the common sense tree has the highest similarity. The calculation method is the bottom of the tree =疋 (TheDeppestLevel) divided by the lowest level of the parentheses tree plus the value of 丨 (the bottom level of the tree I / (the lowest level of the tree + 1)), for example, when the teacher defines " person " and the instructor defines " person " ' Have the same literal, phase Yiyuan and the same English spelling 'after calculation will get a score of 8 / (8 + 1), which means that the two definitions have the highest similarity in this common sense tree, of which 8 is the lowest level of the physical common sense tree, which is defined by the teacher " Teach "quote" and "define" in the tutor, have the same literals, the same meaning, and the same English spelling. After calculation, you will get a score of 1 3 / (1 3 + 1), which means that the two definitions are common sense. The tree has the highest similarity, and its

0213-8982TWF (nl); STLjC-01-B9192; FRANKLIN.ptd $ 6 pages 589549 Five-Explanation (4) — Medium 1 3 is the lowest level of the common sense tree of events, and the definition in the teacher " Education ', the teacher's The definition, education, has the same literals, the same meaning and the same θ ^ guide spelling. After calculation, a score value of 2 / (2 + 1) is obtained, which means that English has the highest similarity in this common sense tree, where 2 defines the bottom layer for the second feature Changma. The fate of the pair is the one with different sub-faces, different meanings, or different spellings in English ^ meaning 'find the two definitions that have the deepest common parent section in the same common sense tree ^ value divided by the bottom of the common sense tree (TheDep pest Level) Add 丨 (the deepest common parent node value / (the lowest level of the common sense tree + Π) to calculate a score value. In addition, if the two are defined as antisense or antisense, add the calculated score The minus sign is a negative scoring value. Member $ 硪 Tree: scoring value (humanl person, human | person, = 8 / (8 + 1); event common sense tree scoring value (teach 13 / (13/1) The first special policy "Sassafras tree: scoring value (educati ο η | education blindness, eduCation 丨 education) = 2 / (2 + 1) d step S18 of the office 'record the scoring value to-record table, will count 2 / (2 + 1) value 'e.g. 8 / (8 + 1)' 13 / 〇3 / 1) and Z / IZ + 1) There are two scoring records in the scoring table δ up to a recording table 'The record table records both the sum of the tree and the event often, for example, 3' (for example, the common sense of the entity is scattered in many places ;: 2 special and two constants, the tree of knowledge) and the goods tree and event that are included in the two words a = total of common sense trees , For example 4 (Zai secret entity often beneficial, wherein the second sense Tree 'flag knowledge tree).

0213-8982TW (nl); STLC-〇i FRANKLIN.ptd page 7 589549 V. Description of the invention (5) Step S20, read the record table to produce a proportional value, the 50 values of the meaning are included in the 4 · 々i record In the table, there is a protagonist * 3 Α Μ and the column is 3 divided by two words. The definition of how much common sense is included in the common sense tree. Second: How much is the definition of the common sense that is discarded: 1. Two words Based on the two values of the information in the record table and the proportional value, divide the sum of the score values in the record table by the sum of the tree with the knife count value in the record table and multiply by the proportional value, such as ^ (13 / (13 + 1)) + (2 / (2 + 1)) / 3) * 3/4 is the average depth of the three common sense trees defined in the common sense tree, where ((8 / (8 + 1) + ( 13 / (13 + 1)) + (2 / (2 + 1)) / 3) is the average of the score values of the three common sense trees, plus the breadth of the definition contained in the common sense tree is the proportional value multiplied by the brackets 1 minus the proportional value, for example (3/4 * (1- (3/4)) produces a similarity, such as 0.80 85 [((8K8 + 1)) + (13 / (13 + 1)) + (2 / (2 + 1)) / 3) * 3/4] + (3/4 * (Bu (3/4)) = 0.8085). (Sum of the scoring values / There are points in the record table Sum of Common Sense Trees * Proportional Value) + (Proportional Value * (1-Proportional Value)) Finally, please refer to Figure 3, which is a schematic diagram showing the calculation concept based on the common sense tree notation according to the embodiment of the present invention Correlation lexicon of the similarity method, step S24. Based on the similarity of the two vocabularies, establish a thesaurus and an antonym, store the calculated similarity of the two vocabularies in the related lexicon, and count each word in the related lexicon and other words. Similarity, build a thesaurus based on words with high similarity, and words based on low similarity

0213-8982TWF (nl); STLC-〇l-B9192; FRANKLIN.ptd page 8 589549 V. Description of the invention (6) The following is an example to explain if the definitions are different or the English spelling is different, two Defined at the same time, when the element has the deepest common parent node value divided by the bracket in the bottom of the common sense tree (TheDeppestLevel) plus 1 (the bottommost common parent node + 1)) to calculate a count Scores, for example, / ,, (Chang Chi tree takes two words, extracts two words from an article, such as poor first ^ first, lifted the staff, according to the structure of the common sense tree to find out what the two words contain _Responsibility. The definition contained in the foreign exchange poverty is, the attribute value ,,,,, etc. bismuth " 1 meaning 'such as the word-Jiu, low, etc. · The sink establishes an antonym thesaurus is located in the value of the Hui poverty The definition is `` attribute value ,, level π, π rich and poor, and, poor '', where `` attribute value '', and `` attribute value '' in the common sense tree, 丨 丨 level, 丨 your heart The temple `` is located in the common sense tree with " poor " and " chip " 2 In the feature common sense tree, the other word rich and rich contains the definition of "attribute value", ", poor", "rich,", "good", where " attribute value " and " rich " are located The value of the mark is ㈣ ±, and "the rich and the poor" is located in the mark common sense tree. It is located in the second special a. The definition of poverty includes: ^ again in the 4 tree ❿ aValue | attribute value-mark value common sense tree rank | Rank-Sign Common Sense Tree LowRank | Low-Sign Common Sense Tree undesired I 莠 — ^ 0 .. • h ^ The 2nd feature common sense tree richness 丨 Negative ~~ Knowledge Common Tree Poor | Poor- & Values Common Sense Tree Definition included in riches: aValue | attribute value ~ do "value common sense tree

0213.8982TWF (nl); STLC-01-B9192; reANKLIN.ptd Page 9 589549 V. Description of the invention (7) richness | rich or poor ~ sign common sense tree rich 丨 rich—sign value common sense tree desired | good-second characteristic common sense tree Next, determine whether the definitions of the two words are located on the same common sense tree. If the two definitions are on the same common sense tree and have the same literals, synonyms, and the same English spelling, the two definitions will be given a common sense tree. The scoring value i has the highest similarity, for example, the definition of the word "poor and rich" and the definition of "economy" " poor and rich "in the common sense tree have the highest similarity. Next, refer to Figure 2 and Figure 2. The diagram is a schematic diagram showing the second feature common sense tree according to the embodiment of the present invention. The operation flow of the embodiment of the present invention will be described below. The definition of the word "poor" is not the same as the word "rich". The same meaning element and different English spellings, so the two commonest parent node values with the deepest definition in the standard common sense tree will be searched. For example, 2, the score value is obtained according to the above score algorithm. Is (2 / (2 + 1)), and because the definition of vocabulary “贱” and the definition of vocabulary ‘good’ are antonyms, the calculated score plus a negative sign produces a negative score. Is-(2 / (2 + 1)). In addition, the definition of poor vocabulary "poor," and the definition of rich and vocabulary "rich" have different literals, different meanings, and different English spellings. Therefore, the two deepest common fathers will be searched in the flag value common sense tree. The node value, for example, 3, according to the above-mentioned scoring algorithm (the deepest common parent node value is the lowest level of the common sense tree + 1)), the scoring value is (3 / (4 + 1)), and due to poor vocabulary The definition of "poor" and the definition of vocabulary "rich π" are antonyms, so the calculated score value plus a negative sign produces a negative score value-(3 / (4 + 1)).

589549 V. Description of the invention (8) Mark and common sense tree: score value (richnessl rich, richness I poor) = 3 / (3 + 1) poor, ri ch I rich) Mark value common sense tree: score (p 〇〇r |-(3 / (4 + 1)) The second feature common sense tree: score value (undesired 丨 match, desired | good) =-(2 / (2 + 1)) > Next, calculate the The score value obtained is recorded in a record table. The total of the common sense tree with the score value contained in the record table is recorded in the record table, for example, 3 (sign common sense tree 'sign value common sense tree, second characteristic common sense tree), and The definition of the two common words in the two words is the sum of common sense trees. For example, it is 3 (sign common sense tree, flag value common sense tree, and second characteristic common sense tree). Sum of common sense trees with a scoring value, for example, 3 divided by the sum of all common sense trees that define how many common sense trees are scattered, for example, a proportional value is generated for 3, such as 3 / 3. Finally, according to the information in the record table and The relationship between the ratio values, calculate the average depth of the two common sense trees defined in the common sense tree and add the common sense tree The breadth of the included definition produces a similarity, such as -0 · 5 丨 6 6. (3 / (3 + 1) + (-3 / (4 + 1) + (-2 / (2 + 1))) / 3) * 3/3 + (3/3 * (1-3 / 3)) =-0 · 5166 According to the above, the method of the present invention for calculating the similarity of concepts according to the common sense tree marking method can correctly express two words The similarity between them and the fast and objective addition of vocabulary rules, so it has great industrial value in the application of knowledge management. Although the present invention has been disclosed as above with preferred embodiments, it is not intended to

Page 11 〇213-8982TWF (nl); STljC-01-B9192; FRANKLIN.ptd 589549 V. Description of the invention (9) The invention is limited, and anyone skilled in the art will not depart from the spirit and scope of the invention. Various modifications and retouching can be made, so the protection scope of the present invention shall be determined by the scope of the attached patent application.

0213-8982TWF (nl); STLC-01-B9192; FRANKLIN.ptd Page 12 589549 Brief description of the diagram Figure 1 is a schematic diagram showing the operation of the method for calculating conceptual similarity according to the common sense tree notation according to the embodiment of the present invention Flowchart; Figure 2 is a schematic diagram showing a second feature common sense tree according to an embodiment of the present invention; Figure 3 is a schematic diagram showing a related thesaurus of a method for calculating conceptual similarity based on a common sense tree notation according to an embodiment of the present invention. [Symbol description] S 1 0 ~ capture two words; S1 2 ~ query the definition of two words according to the common sense tree structure; S1 4 ~ judge whether the definitions contained in the two words are on the same common sense tree; S1 6 ~ When the two definitions are located on the same common sense tree, the score values of the two definitions are calculated; 51 8 ~ Record the score value to a record table; 52 0 ~ Read the record table to generate a proportional value; S2 2 ~ According to the record table The calculation of the relationship between the information and the proportional value produces a similar value; S24 ~ Establish a thesaurus and antonym based on the similarity of the two words; 3 0 ~ related thesaurus.

0213-8982TWF (nl); STLC-01-B9192; FRANKLIN.ptd Page 13

Claims (1)

589549 VI. Application for patent scope 1. * A method for calculating similarity of nasal concepts based on common sense tree marks, which includes the following steps: Provide a common sense tree structure with a plurality of common sense trees; provide a first word and a second word The first vocabulary has at least a first definition and the second vocabulary has at least a second definition; judging whether the first definition and the second definition are located in the same common sense tree, respectively; when the first definition and the first When the two definitions are located in the same common sense tree, at least one score value corresponding to the common sense tree is determined; based on the score value, the similarity between the first word and the second word is determined. 2. The method for calculating conceptual similarity based on the common sense tree notation described in item 1 of the scope of patent application, wherein when the first definition is located in the same common sense tree as the second definition and the first definition is the same as the first The two definitions have the same literals, the same meanings, and the same English spelling. The above-mentioned first definition is the same as the above-mentioned second common sense tree. The score value is the above-mentioned first definition and the above-mentioned second definition. The highest similarity of common sense trees. 3. As described in item 2 of the scope of patent application, the method for calculating the similarity of concepts based on the common sense tree notation, wherein the method for calculating the highest similarity of the common sense tree is (the lowest level of the tree / (the lowest level of the tree) + 1)) ° 4. The method for calculating conceptual similarity based on the common sense tree designation method described in item 1 of the scope of patent application, wherein the steps further include 0213-8982TW (nl); STljC-01-B9192; FRANKLIN.ptd page 14 589549 6. The scope of the patent application is not the same as the second definition above, the meaning is different, or the English spelling is different, which will be in the common sense tree above. Find the deepest common parent node and calculate the score of one of the common sense trees above. 5. The method for calculating conceptual similarity based on the common sense tree notation as described in item 4 of the scope of patent application, wherein the score-value method for calculating one of the above common sense trees is (the deepest common parent node value / (of the common sense tree The lowest level + 1)) 〇6. As described in item 1 of the scope of patent application, the method of calculating conceptual similarity based on common sense tree marking method, which further includes a step of recording the calculated positive and negative score values to a Record Form. 7. The method for calculating conceptual similarity based on the common sense tree marking method described in item 6 of the scope of the patent application, wherein the above-mentioned record table further includes the total T1 of the common sense tree recorded in the record table and the above-mentioned first The sum of the common sense tree T2 of the definition and the second definition above are scattered. 8. The method for calculating conceptual similarity based on common sense tree notation as described in item 7 of the scope of patent application, wherein the above-mentioned T1 value is divided by the above-mentioned T2 to produce a proportional value. 9. The method for calculating the similarity of a concept according to the common sense tree designation method described in item 1 of the scope of patent application, wherein the steps further include scoring the above points when the first definition and the second definition are antisense or opposite. A negative sign is added to the value to produce a negative scoring value. 10. The method for calculating conceptual similarity based on the common sense tree marking method described in item 6 of the scope of patent application, wherein a total of a score value is generated according to the above record table divided by a common sense tree with a score value in the record table Sum 0213-8982TWF (nl); STljC-01-B9192; FRANKLIN.ptd page 15 ~ 'Shenjing patent range ^ The ratio value is multiplied by brackets 丨 minus the above ratio value (the sum of the scoring values / = in the table) The sum of the common sense tree of the scoring values * proportional value) + (proportional value 0-proportional value)) produces a similar value. ~ 11 · As described in Item 10 of the scope of patent application, the method for calculating the similarity of concepts based on the common sense tree mark, wherein the similarity is stored in an association dictionary, and each 3 Phases of other vocabularies = degree, a synonym database is established based on the above-mentioned similar words with a high degree of similarity, and an antonym dictionary is established based on the words with a low similarity degree. Method 1 2-A method for calculating the similarity of nasal outlines according to the common sense tree designation method, which includes the following steps: providing a common sense tree structure; providing a definition of a first word and a second word, the second word having at least The similarity between the first vocabulary and the second vocabulary is determined according to the correlation between the first definition and the second construct. 1 3 · The method for calculating conceptual similarity based on the common sense tree designation as described in item 12 of the scope of patent application, wherein the first definition is true when the first definition and the second definition are located in the same common sense tree. It has the same literals, identical meanings, and same spellings as the second definition above. ”The above-mentioned first definition is the same as the second definition. The highest similarity between the definition and the second definition in the common sense tree. Η. If the U.S. households in the scope of the patent application apply the tree-based labeling based on common sense, the first term above has at least one / number "—definition; and the definition is in the common-sense tree-like structure. 0213-8982TWF (nl); STLC-01-B9192; FRANKLIN.ptd page 16 589549 VI. Method for calculating conceptual similarity by applying the patent scope method, in which the method for calculating the southernmost similarity of the common sense tree is (tree The lowest level of the tree / (the lowest level of the tree + 1)) ° 1 5. The method for calculating the similarity of the concept according to the common sense tree notation as described in item 12 of the scope of patent application, wherein the steps further include the above first definition It is literally different from the second definition above, the meaning is different, or the English spelling is not the same. The deepest common parent node will be found in the common sense tree and a score value of one of the common sense trees will be calculated. 1 6. The method for calculating conceptual similarity based on the common sense tree notation as described in item 15 of the scope of patent application, wherein one of the above common sense trees is calculated as (the deepest common parent node value / (common sense tree) The lowest level + 1)) 〇1 7. The method for calculating conceptual similarity based on the common sense tree marking method described in item 12 of the scope of patent application, wherein the steps further include recording the calculated positive and negative score values To a record sheet. 1 8. The method for calculating conceptual similarity based on the common sense tree marking method described in item 17 of the scope of patent application, wherein the above-mentioned record table further includes the total T1 of the common sense tree recorded in the record table with a score value and the above-mentioned first The sum of the common sense tree T1 of one definition and the second definition above is scattered. 19. The method for calculating conceptual similarity based on common sense tree notation as described in item 18 of the scope of patent application, wherein the above-mentioned Tl · value is divided by the above-mentioned T2 to produce a proportional value. 2. The method for calculating the similarity of a concept based on the common sense tree notation as described in item 12 of the scope of patent application, wherein the steps further include when the first 0213-8982TWF (nl); STLC-01.B9192; FRANKLIN.ptd Page 17 589549 6. The definition of the scope of patent application and the above second definition are antisense or opposite. Add the above score value to a negative sign to generate a Negative score. 2 1. The method for calculating conceptual similarity based on the common sense tree designation method described in item 17 of the scope of patent application, wherein the total of a score value generated according to the above record table is divided by the common sense of the score value in the record table Sum of trees plus-the above scale value multiplied by brackets 1 minus the above scale value (sum of score values / sum of common sense trees with score values in the record table * scale value) + (scale value * (1-scale Value)) produces a similar value. 2 2. The method for calculating the similarity of a concept according to the common sense tree designation method described in item 21 of the scope of patent application, wherein the similarity is stored in a related dictionary, and each word in the related dictionary is counted with other The vocabulary similarity is based on the above high similarity vocabulary to establish a thesaurus, and the low similarity vocabulary is based on the antonym. 0213-8982TWF (nl); STLC-01-B9192; FRANKLIN.ptd Page 18