KR101089338B1 - Method and apparatus for evaluation of original text in bulletin board - Google Patents

Abstract

A method for evaluating the original text in a bulletin board such as the Internet is provided. Extract and vectorize keywords from the text of the original post. Extract keywords from original comments and vectorize them by comments. The similarity between the original keyword vector and the comment vector is compared. Except for a range of comments with low similarity, the rest is selected. A positive keyword and a negative keyword are extracted from the selected comments to obtain a positive score and a negative score for each comment. Scores are given to the text of the post based on the scores for the pros and cons.

Description

Method and apparatus for evaluating the original text of the post {Method and apparatus for evaluation of original text in bulletin board}

Embodiments of the present invention relate to a method and apparatus for evaluating a text of a post, such as the Internet, by analyzing comments of a post posted on the Internet.

In times when the Internet was not active, it was difficult to get data on a particular subject. And in the early days of the development of the Internet, it was relatively easy for people with similar Internet interests to gather together in a virtual space on the Internet. Although the amount of data was small, finding data on a particular area of the Internet was relatively easy.

On the other hand, today, the consciousness and use of various industries and individuals on the Internet has changed from the past, and a vast amount of materials are disclosed on the Internet.

It is convenient to be able to use such a large amount of data, but on the other hand, it is more difficult to find out which of these many materials are useful and which are not useful. In other words, in the past, a small number of data became a problem, but nowadays, it is more important to select a stone from a large amount of data. In other words, evaluating various materials has become an important place due to the rapid development of the Internet and the increase of information volume.

An embodiment of the present invention is to provide a method and apparatus for evaluating a text from a comment on a post posted on the Internet.

According to an embodiment of the present invention, a text keyword vectorizing step of extracting and vectorizing a keyword from a text of a text of a post text for a post in a telecommunication network by the text keyword vectorizing means; A comment keyword vectorizing step of extracting keywords from n comments (n is a natural number of 1 or more) for the original text by the comment keyword vectorizing means and vectorizing each comment; A vector similarity comparison step of comparing, by vector similarity comparing means, similarities between the original keyword vector and the n comment keyword vectors; A comment selecting step of selecting, by the comment selecting means, only the remaining comments except for a predetermined range of low similarity according to the ranking of the similarity or the similarity number; Obtaining, by the affirmative and disagreeing score obtaining means, affirmative and disagreeing score obtaining points for each comment by extracting affirmative keywords and disagreeing keywords from the selected comments; And an original text scoring method for assigning a score to the original text based on the positive score and the opposite score by the original score assigning means.

According to an embodiment of the present invention, for a post in a telecommunications network, original keyword vectorization means for extracting and vectorizing a keyword from the text of the original post; A comment keyword vectorization means for extracting keywords from n comments (n is a natural number of 1 or more) for the original text of the post and vectorizing each comment; Vector similarity comparison means for comparing the similarity between the original keyword vector and the n comment keyword vectors; Comment selecting means for selecting only the remaining comments except for a predetermined range of low similarity according to the ranking of similarity or similarity value; A means for acquiring and disagreeing scores for extracting the disapproval and disapproval keywords from the selected comments to obtain a score for each comment; And an original text scoring means for assigning a score to the original text based on the affirmative score and the opposite score.

The embodiment of the present invention evaluates the comments on the posts posted on the Internet, and evaluates the original text by combining the evaluations of the plurality of comments.

According to an embodiment of the present invention, even if a separate voting function is not added to the bulletin board, the evaluation of the original text is made with only ordinary comments.

Alternatively, a detailed evaluation may be made according to the degree of affirmation (agreement) or disagreement (a favorability) in the text, so that a more detailed evaluation may be made compared to the voting function of only one vote or affirmative vote.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. However, this is only an example and the present invention is not limited thereto.

In describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intention or custom of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

The technical spirit of the present invention is determined by the claims, and the following embodiments are merely means for efficiently explaining the technical spirit of the present invention to those skilled in the art.

1 is a flowchart illustrating a text evaluation method according to an embodiment of the present invention.

Referring to FIG. 1, a keyword and a frequency are extracted from the text of a post written on an Internet bulletin board or the like in step 102. The text is all information in the form of a letter, which may mean the text of the title and the content of the original text, and may further be modified such as automatic recognition of the text in the image file.

In extracting keywords, a keyword dictionary and stemming analysis can be used. For example, morphological analysis can be used to eliminate the effects of unnecessary irradiation (s, s, s). In other words, not all texts are keywords, and morphological analysis can be used to find keyword candidates. Morphological is the smallest unit of meaningful language units. In morphological analysis, index terms are identified by identifying real morphemes that have substantial meanings. Used for extraction. Parts of speech used as index terms are mostly nouns (or verbs), so they are identified through morphological analysis, and in the case of irregular use, dropout, or abbreviation, they are processed to find a prototype. Programs that automatically perform stemming and keyword extraction are being used.

In morpheme analysis, for example, in the case of Hangul, the text consisting of 'noun + search' may be divided into nouns and surveys, and the keyword candidate may be found in nouns other than the survey. The keyword candidate refers to a word that has not yet been selected as a keyword and has been verified to be a keyword.

In addition, the Internet was mentioned earlier, but this is an example and should be understood as a telecommunication network (electronic information communication network) in which a computer is involved. Therefore, even if the intranet is not connected to the outside it is obvious that the basic idea of the embodiment of the present invention can be applied if only a certain amount of scale. Therefore, it should be seen that the present invention is applied to an entire electronic information communication network through electronic means. In addition, it is a matter of course that the communication can be applied by wire or wireless.

The keyword extracted from the article (original text) (including information on the frequency of the keyword, if necessary) is referred to as keyword set A. The extracted keywords and / or frequencies are vectorized, which is referred to as the original vector VA.

Next, in step 104, a comment on the original text of step 102 is analyzed. In other words, the keyword is extracted from the comment on the original text of the post.

A comment, as described above, will be the user's response after viewing the original publisher's description. These comments may of course be plural and may be by multiple users.

In the keyword extraction process, a morphological analysis method may be used. Keyword dictionaries may also be used. Assuming n comments, the keyword set of each comment may be referred to as B1, B2, B3, ..., Bn. From each of these, the vectors VB1, VB2, VB3, ..., VBn of the comment can be obtained.

In extracting the keyword while performing step 102 and / or step 104, it is desirable to exclude banned words or stop words. Banned words exclude abusive language and advertising. In addition, the comment containing the banned words may be excluded from the evaluation or given the lowest score, and the ID of the person who habitually performs abusive or post posting may be managed in a manner of excluding these comments from the evaluation. Also, stopwords are words that have no value as index words, and are excluded after being set in advance. Kind words and stopwords may be excluded after the keyword extraction process, or may be excluded from the keyword extraction process, but the latter would be preferable in terms of efficiency (efficient use of computing resources).

For example, terms used in abusive language and advertising phrases can be managed as separate dictionaries to prevent them from being selected as keyword candidates. Also, stop words that are not of value as index terms can be managed as separate dictionaries as keyword candidates. It can also be chosen.

Next, in step 106, the similarity of the original vector and each comment vector is compared.

The original vector refers to VA obtained in step 102, and each comment vector refers to VBk (k is a natural number from 1 to n) obtained in step 104. That is, VA and VB1 are compared, VA and VB2 are compared, VA and VB3 are compared, ..., and VA and VBn are compared to calculate the similarity in each case.

At this time, there are various techniques as a comparison method of similarity. Representative examples may include (1) a keyword comparison method of a Boolean model, (2) similarity calculation using Euclidean distance formula, and (3) similarity calculation using cosine formula. This is only an example and any method that can compare the similarity may be used.

The keyword set is converted to a vector to perform a search using a vector. Of course, the keyword set is not necessarily converted to a vector to be searched.

Some of these are described below.

The similarity measure by Euclidean distance, that is, the Euclidean distance formula, is to calculate the distance between two points in multidimensional space.The two points (vector) are defined as (p1, p2, p3, p4, ...) and (q1). , q2, q3, q4, ...), the Euclidean distance formula is:

On the other hand, the similarity measurement by the cosine formula is as follows.

The cosine formula is used to calculate the cosine value between vectors. At this time, the larger the cosine value, the higher the similarity.

Next, in step 108, the comments are ranked from the highest ones (eg, in descending order) by the similarity with the original text to give the rankings to the comments.

That is, among the similarity between VA and VB1, the similarity between VA and VB2, the similarity between VA and VB3, ..., and the similarity between VA and VBn, the highest and highest are ranked.

Next, in step 110, only comments whose similarity is equal to or greater than a predetermined predetermined value are selected. For example, the similarity between VA and VB1, the similarity between VA and VB2, the similarity between VA and VB3, ..., the similarity between VA and VBn, respectively, between 0 and 1 (e.g., the closer to 1, the higher the similarity). Assumptions) is 0.4, 0.3, 0.1, 0.4, 0.7, 0.5, ... 0.9 and if a predetermined predetermined value (i.e. the threshold of similarity) is 0.35, a comment having a similarity of 0.35 or more among the above values Only the fields are ranked. Such low similarity comments can be excluded from subsequent evaluation because they have little relationship with the original content. In addition, the term "comment" used in relation to the evaluation of the comment and the original text of the post means a similar comment selected in this process.

With respect to steps 108 and 110, in step 108 the ranking is ranked according to the similarity, and in step 110 the thresholds of the similarity numbers are sorted accordingly, although only one of these judgment measures may be used if necessary. That is, an embodiment of selecting only comments above a certain rank by calculating a rank or selecting only comments above a certain number by calculating a similarity value may be possible.

Next, in step 112, the selected commenter may be given an incentive score based on the similarity value. The incentive value may be used as the similarity value itself or by processing the similarity value (for example, multiplied by a predetermined weight). This score may be aggregated over time or aggregated to determine the commenter ranking. The ranking of the commenter at this time indicates the ranking of the faithful comments related to the original text.

For example, among n comments, if a comment 1, a comment 4, and a comment 5 have similarities exceeding a predetermined threshold, and the values are 0.4, 0.45, and 0.7, respectively, the similarity values of 0.4, 0.45, and 0.7 ( s) itself may be given as an incentive score, or a value obtained by multiplying a predetermined weight w1, that is, s × w1, may be given as an incentive score. w1 can be made into 0.1, for example.

Such evaluation may end with an evaluation of a comment on one original text, or, as described above, may be aggregated by user (identified by ID) and used as a statistic. Cumulative eggs can be processed in various ways, including not only the cumulative total of the original text, but also the cumulative total of the original text of the entire bulletin board, and the cumulative total of the original text of the entire bulletin board.

Next, in step 114, the affirmative keyword and the opposite keyword are extracted from the selected comments. At this time, the above-described morphological analysis and keyword dictionary may be used. It's also a good idea to find out the frequency of the pros and cons related keywords. In addition, it is preferable to use an emphasis table (to be described later) and an antonym dictionary (to be described later) together. The terms 'yes' and 'opposition' are not limited to agreements and disagreements. Agree is a general term for a term that expresses empathy in the text, such as approval, consent, support, approve, permission, permission, acceptance, favour, friendship, good, in a broad sense, and vice versa. Therefore, the present invention can be applied to various cases according to a method of determining the meaning of pros and cons.

If the frequency exceeds the appropriate number (for example, three times), only the frequency set in advance is set. For example, a comment might contain the phrase "yes" and "yes", but if you say "yes" six times and give it six times as many points, then all users will be able to comment You may repeat many times (eg 100 times, 1000 times), which is undesirable. Therefore, in this case, the proper frequency set in advance is set to 3, and the meaning is given only up to 3 even if the word 'yes' is 6 times. This will limit attempts to increase the number of pros and cons by using too many pros and cons keywords.

In addition, if necessary, among the modifiers of the affirmative keyword and the opposite keyword, the emphasis modifier can be found by a morphological analysis technique. For example, 'highly', 'strongly', 'strongly' and 'highly' are the emphasis modifiers. According to this configuration, 'yes' and 'very positive' can be treated differently.

To do this, you can use the pros and dictionaries. Pros and dictionaries contain words that are used to mean pros (for example, affirmative, consent, support, sympathy, etc.). ) Is included. On the other hand, the matter regarding the above-described emphasis word may be processed by the emphasis word table having a (strength word, weight) field.

In addition, it is preferable to use an antonym dictionary, which can be composed of (word, antonym) fields, which is necessary to determine the phrase 'I do not agree' with a disagreement (ie, vice versa). . Without this antonym, the phrase 'don't agree' could simply be regarded as 'agreement' by morphological analysis. Of course, you can judge the phrase “do not disagree” with the intention of opposing (ie, consent).

It is also possible to use a synonym dictionary instead of the antonym dictionary. The synonym dictionary may be composed of (words, synonyms) fields, and for example, 'don't agree' may be set as 'opposite' synonyms. In addition, the words in the antonym and thesaurus can be compound words, phrases, clauses, etc. consisting of a plurality of words as well as words.

Through this process, each player's comment is given a positive and negative score.

The large framework for performing step 114 (i.e., use of the agree dictionary, the antonym dictionary, the emphasis table, and the antonym dictionary) has been described. Various implementation manners for step 114 will be described later in more detail with reference to FIGS. 2, 3, and 4.

Next, in step 116, various measures can be obtained for the text of the post from the positive and negative scores of each comment obtained in the previous step. The various measured values herein may be any values processed using the values obtained in the above-described steps, for example, as follows. This is an example, and it is also possible to define and use an item not defined below or to use a slightly modified definition of each item below. Comments in the following scale refer to similar comments selected for evaluation as described previously.

* Number of comments: The number of comments with a positive score greater than a certain number (e1 ≥ 0).

* Number of negative comments: Number of comments with a negative score greater than a positive score (e1 ≥ 0)

* Comment ranking in favor (in order of score)

Ranking of comments in reverse order

* Agree: Proportion of comments in comments

* Relative rate: The percentage of negative comments in the comment

* Withdrawal Rate: 1-Yes-No

* Total score in favor: Total score in favor of comments

* Total score against: Total score of the score in the comments

* Average approval score of pros: Aavg

* Average Opposite Score for Opposite Comments: Davg

* Total sum of (Yes-Negative scores) for each comment: Ptotal

→ If Ptotal is above a certain value (e2 ≥ 0), it is determined that the post is larger than the opposite.

* Overall Average Of Each Comment's (Pros-Opposite Scores): Pavg

→ If Pavg is above a certain value (e3 ≥ 0), it is determined that the post is larger than the opposite.

* Weighted the overall average (Pavg) of each comment's (favor scores-dissent scores) by the number of comments: Pavg × Number of pros ×

¡Æ where c1 is an adjustment factor and c1 may be given a different weight for each number of pros. For reference, if c1 is 1, it is equal to Ptotal.

* Ptotal × c2

¡Æ where c2 is an adjustment factor and c2 may be assigned a different weight for each Ptotal score band. For reference, 1 is equal to Ptotal.

Next, in step 118, appropriate measurements can be selected (eg, Pavg and Ptotal) from the measurements in step 116 and assigned as the score of the original text.

Although it is said that a suitable measurement is selected from the measurements in step 116, it would be reasonable to think of it as a concept that includes only the measurements necessary in step 116 and then use in step 118.

Next, in step 120, ranking may be given in order of the score of the post. That is, the ranking is given to the posts by sorting the highest score in descending order according to the measurement. For example, if you use Pavg, you can give the ranking by sorting in descending order from the highest score of each comment minus the opposite score.

Next, in step 122, an incentive may be given to the original publisher by the score given to the original. The incentive score may be, for example, a value using a given score Pavg or a processed value (Pavg × w 2, where w 2 is an adjustment factor). This score can be accumulated over time or in aggregate to calculate publisher ranking. At this time, the ranking of the publisher will represent the ranking obtained by consent from many comments.

On the other hand, in practicing the present invention, it may be considered unnecessary to give a ranking in the above-described steps 108 and 120 as necessary. That is, if the similarity ranking of comments and / or the ranking of the original text is unnecessary and only the similarity number of the comments and / or the score value of the original text are needed, the ranking given in steps 108 and 120 may be omitted, and the present invention may be implemented only by the numerical values. It may be.

2 illustrates an example of an implementation of step 114 of FIG. 1 described above.

As described above, step 114 is a process of obtaining a score for each comment, and the process of obtaining such a score may be implemented in various ways. Hereinafter, one embodiment of such various methods will be described.

In summary of the step 114 of FIG. Extracting the stress modifiers), thereby obtaining a score for each comment.

As a concrete implementation manner thereof, in step 114-2, each field of the proactive dictionary is composed of (a pronoun), and each field of the opposite dictionary is composed of (an opposite). This is the simplest implementation of the pro and dictionaries.

In step 114-2, the agree dictionary and the opposite dictionary must be set in advance. In addition, the antonym (or synonym) dictionary must also be set in advance. The antonym dictionary is a dictionary for accurately recognizing “not agree” as described above. If the antonym dictionary uses (words, antonyms) as a field, the synonym dictionary uses (words, synonyms) as a field. Therefore, the same effect can be obtained by using either an antonym dictionary or a synonym dictionary. In the following description, the antonym dictionary is used for convenience, but the synonym dictionary may be used as a matter of course.

Next, referring to step 114-4, by morphological analysis, the number of pros and cons for each comment is determined. In step 114-6, if the number of pros in a particular comment is greater than the number of converses. The comment is considered to be in favor, and in the opposite case, the comment is denied.

In this way, comments that use more of the favor keywords than the opposite keywords are considered as affirmative, and comments that use more than the affirmative keywords can be reversed. On the other hand, comments that do not use the pros and cons keywords at all may be treated as absent.

In other words, this method has the effect of voting one vote per comment. Through this, it is possible to calculate the number of pros, the number of dissent, the rate of approval (a ratio of the like comments among similar comments), the rate of disagreement (the ratio of disagree comments among similar comments), and the rate of abstention. This approach may be compared to automatically performing the existing voting method. Of course, in the existing voting method, since the functions related to voting must be implemented in the bulletin board, it will not be exactly the same.

3 illustrates another example of an implementation of step 114 of FIG. 1 described above.

As described above, step 114 is a process of obtaining a score for each comment, and the process of obtaining such a score may be implemented in various ways. Hereinafter, one embodiment different from those described in FIG. 2 among various methods will be described.

In summary of the step 114 of FIG. Extracting the stress modifiers), thereby obtaining a score for each comment. The antonym dictionary is also used.

As a specific implementation manner thereof, first, in step 114'-2, a proactive dictionary composed of a (positive word, a positive score) field, and a reverse dictionary consisting of a (anti-correlation, negative score) field are used. Here, each score has a value between 0 and 1. For example, 0.6 points for 'OK', 0.8 points for 'yes', and 1.0 points for 'support' could be assigned. This can be determined according to the strength of the strength of the dictionary.

That is, in step 114'-2, the affirmative dictionary and the antonym dictionary are used, for example, the affirmative dictionary is composed of the (agreement, affirmative score) field, and the antonym dictionary is a (counter negative, negative score) field. It is preferable to do so in advance since each component (that is, the creation of such a dictionary) to be included in the field corresponds to a basic setting.

In addition, it is preferable that the emphasis word table composed of the (strength word, weight) field is also set in advance and used together. The antonym dictionary is also preset and used together in step 114'-2. The antonym dictionary is a dictionary for accurately recognizing “not agree” as described above. In other words, an antonym dictionary can be used to convert "don't agree" to "opposite" keywords. If the antonym dictionary uses (words, antonyms) as a field, the synonym dictionary uses (words, synonyms) as a field. Therefore, the same effect can be obtained by using either an antonym dictionary or a synonym dictionary.

Next, in step 114'-4, keywords and emphasis words of the comments are extracted together. The keywords include the pros and cons, and thus the extracted ones may be the pros, the opposites, other keywords, and the emphasis. Keywords are converted using an antonym or a synonym dictionary during or after keyword extraction. For example, convert "I don't agree" to the "opposite" keyword. Of course, the accent qualifier can also be used, so you can convert ‘very disagree’ to ‘very disagree’.

Among these, explanations will be made regarding pros and cons. Focusing on the affirmative words, if there were no emphasis words, they would be extracted as the affirmative words alone. In the case of the apology alone, a predetermined affirmative score may be given, and a weighted score may be given to the apology + emphasis (that is, when both the apology and the emphasis are extracted). The sum of each of these adds up to the total score for the pros. Of course, the same operation can be performed also for the opposite language.

On the other hand, when the weighting by the emphasis is described in more detail, the weighting score is calculated using the emphasis table composed of the (strength, weight) field. This is for score weighting by emphasis. For example, 'absolute' may be a weight of 1.0, 'very' may be a weight of 0.6, and 'little' may be 0.3.

Specific calculation method is the following formula

[Word score + {word score × (weight-w3)}] / (word maximum score)

It can be by.

w3 is an adjustment factor and has a value between 0 and 1, and may be, for example, 0.5. The reason for the weighting-w3 (for example, w3 is 0.5) is to reduce the score of the accent word that weakens the meaning of a little, a little, etc. In addition, the 'maximum score of words' in the denominator of Equation 3 is the maximum value that the molecular part of the equation can have, and therefore, the result of the equation is normalized to have a value between 0 and 1. It is also possible to modify the above equation. For example, an emphasis word that weakens meaning may have a weight of a negative integer. In this case, the above equation can be modified to "[word score + {word score x weight}] / (word maximum score)".

When reflecting the weights, the total score of the pros (Td) for each comment is the sum of all scores (including weights) of the pros in the comment, and the total score of the negatives (Td) is the sum of all the scores of the opposite word in the comment. Sum (including weights). That is, the total score of the pros is the sum of the 'favor scores' of each approve word and the 'proscore scores that are weighted by the accent words of the proximal words combined with each accented word' for the appearance of all the pros in the comment. The total score of the opposite word is the sum of the 'counter word scores' of each of the opposite word words and the 'counter weights of the opposite words combined with the emphasis' for each occurrence of the opposite word in the comment.

Next, in steps 114'-6, the total scores for the negative and negative words of each comment are normalized. That is, it is possible to convert the total scores of the pros and cons of each comment to have a value between 0 (low) and 1 (high).

For example, this is possible by the following formula. Let NTa be the normalized pros and cons and NTd.

NTa = Ta / (maximum number of pros)

NTd = Td / (maximum counter total)

Here, the maximum value of the total number of approved words is the maximum total number of approved words that can be given to the comment, and the maximum total number of negative words is the maximum total number of negative words that can be given to the comment.

4 illustrates another example of an implementation of step 114 of FIG. 1 described above.

As described above, step 114 is a process of obtaining a score for each comment, and the process of obtaining such a score may be implemented in various ways. Hereinafter, an embodiment different from those described with reference to FIGS. 2 and 3 will be described.

In summary of the step 114 of FIG. Extracting the stress modifiers), thereby obtaining a score for each comment. Also, in this process, keywords can be converted using an antonym dictionary or a thesaurus.

As a specific implementation of this, referring to step 114 "-2, a proactive dictionary consisting of the (agreement, the emphasis, the affirmative score) field, and a counter dictionary consisting of the (antonym, the emphasis, the negative score) field are used. This is different from that described in Fig. 3 in that the fields constituting the proactive and thesaurus are already included in the accent dictionary, and the antonym dictionary or the synonym field composed of the (word, antonym) field. Synonym dictionaries. Here, in step 114 "-2, it can be seen that the basic settings of the proactive dictionaries and the opposite dictionaries themselves do not necessarily have to be done at this time. That is, it is preferable to make such a dictionary in advance because it corresponds to a basic setting. The antonym dictionary (or synonym dictionary) must also be set up in advance.

Yes and Negative scores are given when words are used with emphasis. If no emphasis is used, the original word score is given. At this time, each score should be between 0 and 1.

Step 114 " -4 is then performed, which can be performed similarly or identically to step 114'-6 described above in connection with FIG.

In other words, the embodiment of FIG. 4 and the embodiment of FIG. 3 will be different in that the emphasis word field and the opposite word dictionary are included or a separate emphasis dictionary is created.

5 briefly describes the method described above as an example of an actual post.

Looking at Figure 5, it can be seen that the original text in the box and the user's comment on it depends on 10. In the example of FIG. 5, the ID of the person who has commented on each comment is displayed to identify who has commented. For example, comment 4 and comment 6 are written by the same person.

In the example of FIG. 5, the comment is illustrated as one or two short sentences, but a formal post (for example, 10 to 20 sentences in one comment) may be created for the original text. However, in FIG. 5, only short texts and comments are given as examples for simplicity of explanation.

The embodiment of FIG. 5 will be briefly analyzed according to the steps of FIG. 1.

According to step 102, the original vector may be selected as a living organism, animal, plant, seed plant, seed plant, genus plant, monocotyledonous plant, dicotyledonous plant, or the like.

Next, according to step 104, the vectors 1 to 10 of comments (VB1 to VB10, respectively) may be selected.

The vectors obtained in steps 102 and 104 may be the result of the morphological analysis and the stopword / stopword filtering already applied.

According to step 106, the original vector VA and the comment vectors VB1 to VB10 are compared, respectively.

According to step 108, the similarity ranking is ranked. For example, let's say that the ranking is set in order of comment 5, comment 2, comment 8, comment 9, comment 1, comment 10, comment 6, comment 3, comment 4 and comment 7 is excluded from the ranking.

For example, comments 3 and 4 have a positive or negative intention, but the ranking is low because there is no related word in the original keyword. However, in the operation of the system, the pros and cons may be registered as relevant. In addition, comment 7 has a lot of banned words can be seen as excluded from the ranking.

According to step 110, comments 3 and 4 having a low similarity ranking are excluded according to a preset criterion. Of course, in the case of comment 7, it was already excluded before judging similarity because there were many banned words, but even if it was not excluded due to banned words, it is likely to be excluded because of low similarity.

In accordance with step 112, the creator of the comment is awarded an incentive score. This score may be accumulated by user.

According to the step 114, a positive score and a negative score are appropriately calculated among the high ranking comment 5, comment 2, comment 8, comment 9, comment 1, comment 10, and comment 6. Step 114 may be performed in various ways, for example, in the manner shown in FIGS.

For example, the 'consent' appears 4 times in the comment 2, but it can be regarded as only 1 or 2 times and scored. The same is true for comment 10, which appears five times. In addition, it is possible to understand 'not agree' in Comment 6 as the opposite meaning, not the consent, by the above-mentioned counter- dictionary. You can also give weight to comment 8 with the word `` very very '' as the emphasis. Of course, in some embodiments, the emphasis word may be ignored.

In accordance with steps 116 and 118, measurements are obtained for the text based on the pros and cons of the comments. According to step 120, a ranking is given to the post (the original). In FIG. 5, there is only one original text (an article in a square box), but assume that there are more posts. For example, there are five other texts, and their scores (comparison scores) are compared to show that the text with the highest score is the highest ranking.

According to step 122, an incentive score may be given to the publisher of the original text of FIG. 5.

In the example of FIG. 5, a simple sentence is taken as an example, but as the sentence becomes longer and the number of samples increases, more detailed classification and accurate evaluation will be possible.

6 shows an apparatus according to an embodiment of the invention.

In FIG. 6, the device 600 of the present invention is connected to a telecommunication network (electronic information communication network) represented by the Internet. The telecommunication network may be the Internet or an intranet, regardless of whether it is wired or wireless or not.

The apparatus 600 of the present invention includes a text keyword vectorizing means 602, a comment keyword vectorizing means 604, a vector similarity comparing means 606, a comment selecting means 608, a positive score / inverse score obtaining means 610. , And text score assigning means 612.

The original keyword vectorizing means 602 extracts and vectorizes a keyword from the text of the original text of the post in the telecommunication network.

The comment keyword vectorizing means 604 extracts a keyword from n comments (n is a natural number of 1 or more) for the original text of the post and vectorizes each comment.

The vector similarity comparison means 606 compares the similarity between the original keyword vector and the n comment keyword vectors.

The comment selecting means 608 selects only the remaining comments except for a range of comments having a low similarity according to the ranking or similarity value of the similarity.

The affirmative score / inverse score obtaining means 610 extracts affirmative keywords and opposite keywords from the selected comments to obtain a score for each comment.

 The original text scoring means 612 assigns a score to the original text based on the positive score and the opposite score.

An example has been described above. For example, the Internet, but it is an example and should be understood as a telecommunications network in which computers are involved. Therefore, even if the intranet is not connected to the outside it is obvious that the basic idea of the embodiment of the present invention can be applied if only a certain amount of scale. Therefore, it should be seen that the present invention is applied to an entire electronic information communication network through electronic means. In addition, it is a matter of course that the communication can be applied by wire or wireless.

In addition, for example, an antonym dictionary was used to avoid misinterpretation of the opposite's intention as an intention of consent and to avoid misunderstanding the intention of the opposite. It is self-evident. In other words, even if an antonym dictionary is mentioned in a claim, it is clear that the synonym dictionary is the same technical idea.

Meanwhile, an embodiment of the present invention may include a computer readable recording medium including a program for performing the methods described herein on a computer. The computer-readable recording medium may include program instructions, local data files, local data structures, etc. alone or in combination. The media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CD-ROMs, DVDs, magnetic-optical media such as floppy disks, and ROM, RAM, flash memory, and the like. Hardware devices specifically configured to store and execute program instructions are included. Examples of program instructions may include high-level language code that can be executed by a computer using an interpreter as well as machine code such as produced by a compiler.

Although the present invention has been described in detail with reference to exemplary embodiments above, those skilled in the art to which the present invention pertains can make various modifications to the above-described embodiments without departing from the scope of the present invention. Will understand.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims below and equivalents thereof.

1 is a flowchart illustrating a text evaluation method according to an embodiment of the present invention.

2 illustrates an example of an implementation of step 114 of FIG. 1 described above.

3 illustrates another example of an implementation of step 114 of FIG. 1 described above.

4 illustrates another example of an implementation of step 114 of FIG. 1 described above.

5 briefly describes the method described above as an example of an actual post.

6 shows an apparatus according to an embodiment of the invention.

Claims (17)

An original keyword vectorizing step of extracting and vectorizing a keyword from the text of the original text of the post in the telecommunication network by the original keyword vectorizing means; A comment keyword vectorizing step of extracting keywords from n comments (n is a natural number of 1 or more) for the original text by the comment keyword vectorizing means and vectorizing each comment; A vector similarity comparison step of comparing, by vector similarity comparing means, similarities between the original keyword vector and the n comment keyword vectors; A comment selecting step of selecting, by the comment selecting means, a range of comments having a high similarity according to the ranking or similarity value of the compared similarities; By the affirmative and disagreeing score acquiring means, extracting the affirmative keyword and the disagreeing keyword from the selected comments, and calculating a score for each comment according to the number of the affirmative keyword and the disagreeing keyword, and agreeing the score and the disagreeing score. Acquiring the pros and cons scores; And And a text scoring step of assigning a score to the text of the post based on the positive score and the opposite score, by the text scoring unit. The method of claim 1, The comment selection step, Ranking of the comment according to the ranking of the similarity, the original text evaluation method, characterized in that excludes the comment with a low similarity ranking. The method of claim 1, The comment selection step, According to the numerical value of similarity, the original text evaluation method, characterized in that the comment with a low similarity value is excluded. The method according to claim 2 or 3, The comment selection step, in addition, Based on the similarity value or rank, the original text evaluation method characterized in that the comment author is given a predetermined incentive score. The method of claim 1, Between the pros and cons score obtaining step and the original score assigning step, Further obtaining a measure of the original text from the pros and cons of each comment; The measurement value, * The number of positive comments (i.e., the number of comments with a certain number (e1) greater than the negative score) * The number of negative comments (i.e., the number of comments where the negative score is a certain number (e1) more than the positive score) * Comment ranking in favor (in order of score) Ranking of comments in reverse order * Affirmative (i.e., the percentage of negative comments in the comment) * Negative rate (i.e., the percentage of negative comments in the comment) * Abstention rate (ie 1-Yes-No) * Total score for the vote (ie total score of the vote in the comments) The total score of the opposite score (ie the total score of the negative score in the comments); * Average approval score of pros * Average dissent score of dissent comments * The total sum of each vote's (Yes-Negative scores) * Overall average of each vote's (favorite score-disagree score) * Weighted by the number of pros and cons to the overall average (Pavg) of (Pros-Disagree) for each comment. Is one or more of In the original text scoring step, the original text evaluation method characterized in that the score is assigned to the original text using the measurement. The method according to any one of claims 1 to 3, After the text scoring step, Ranking the posts by the score of the original text of the post; And Giving a predetermined incentive score to the original publisher by the score of the original text of the post Post text evaluation method that includes more. The method according to any one of claims 1 to 3, After the text scoring step, Giving a predetermined incentive score to the original publisher by the score of the original text of the post Post text evaluation method that includes more. The method of claim 1, The acknowledgment point and the opposite point obtaining step, The number of pros for each comment by morpheme analysis, using a prologue dictionary containing the (favorite) field, an antonym dictionary containing the (antonym) field, and an antonym dictionary containing the (word, antonym) field. And identifying the number of opposites, wherein the antonym dictionary includes the steps of: preventing the intention of the opposite opinion to be regarded as the intention of consent or the intention of the consent to be regarded as the opposite intention. And If the number of pros in a particular comment is greater than the number of negatives, the comment is considered to be approved, and if the opposite is the case, the comment is denied. Original text evaluation method comprising a. The method of claim 1, The acknowledgment point and the opposite point obtaining step, Includes a dictionary of apologies that include the Pros and Cons scores field, a counterpart dictionary that includes the Counter, Negative Score field, and a weight table consisting of the Highlight, Weight field, and the Word, Antonyms field. Using an antonym dictionary, the total score of the pros and cons is the sum of the 'favor scores' of each word and the 'favor scores reflecting the weights of the accents'. It is a sum of 'antonym score reflecting weight by accent word', but the antonym dictionary is characterized in that the opposite expression of intention is not regarded as the intention of consent or the intention of consent is not regarded as the opposite intention. How to rate the text of the post. The method of claim 1, In the pros and cons score acquisition step, YES using a proactive dictionary that includes the (accept, accent, and affirmative) field, an antonym dictionary that includes the (antonym, an accent, a negative score) field, and an antonym that includes the (word, antonym) field. Compute the total score of the word and the total score of the opposite word, wherein the antonym dictionary checks that the intention of disagreement is not regarded as the intention of consent or the intention of consent is not regarded as the intention of disagreement. . The method according to claim 9 and 10, The weight by the emphasis is [Word score + {word score × (weight-w3)}] / (word maximum score) Reflected by The w3 has a value between 0 and 1 as an adjustment factor, and the maximum score of a word is a maximum value that a molecular part of the equation can have. The method according to claim 9 and 10, The total score (Ta) in favor of the comment and the total score (Td) in the opposite words of each comment are as follows. NTa = Ta / (maximum number of pros) NTd = Td / (maximum counter total) NTa is the normalized total number of approved pros, NTd is the normalized negative total, the maximum total number of pros is the maximum total number of pros that can be given to a comment, and the maximum total of negatives can be given the maximum Original text evaluation method characterized in that the total score is the opposite. A computer-readable recording medium having recorded thereon a program for performing the method according to any one of claims 1 to 3, 5 and 8 to 10 on a computer. Text keyword vectorizing means for extracting and vectorizing the keyword from the text of the text of the post for the post in the telecommunication network; Comment keyword vectorizing means for extracting keywords from n comments (n is a natural number of 1 or more) for the original text of the post and vectorizing each comment; Vector similarity comparison means for comparing the similarity between the original keyword vector and the n comment keyword vectors; Comment selecting means for selecting a predetermined range of comments having a high similarity according to the ranking of similarity or the similarity value; Affirmative and disagreeing score obtaining means for extracting affirmative and disagreeing keywords from the selected comments, and calculating affirmative and disagreeing scores for each comment according to the number of extracted and disagreeing keywords; And And an original text scoring means for assigning a score to the original text based on the positive score and the opposite score. The method of claim 14, The pros and cons score obtaining means, The number of pros for each comment by morpheme analysis, using a prologue dictionary containing the (favorite) field, an antonym dictionary containing the (antonym) field, and an antonym dictionary containing the (word, antonym) field. And identifying the number of antonyms, wherein the antonym dictionary comprises: means for the contrary intention to be regarded as an intention of consent or for the intention of consent not to be regarded as an intention of intention; And Means that a comment is considered to be approved if the number of approved words in the comment is greater than the number of negative words, and if it is reversed, the comment is considered as disapproval. Original text evaluation device comprising a. The method of claim 14, The pros and cons score obtaining means, Includes a dictionary of apologies that include the Pros and Cons scores field, a counterpart dictionary that includes the Counter, Negative Score field, a weight table that includes the Highlight, Weights field, and the Word, Antonyms field. Using an antonym dictionary, the total score of the pros and cons is the sum of the 'favor scores' of each word and the 'favor scores reflecting the weights of the accents', and the total scores of the opposite words are the 'counter words of each word' and It is a sum of "antonym score reflecting the weight by emphasis", but the antonym dictionary is characterized in that the opposite expression of intention is not regarded as the intention of consent or the intention of consent is not regarded as the opposite intention. Post text evaluation device. The method of claim 14, In the means for obtaining the positive and negative scores, YES using a proactive dictionary that includes the (accept, accent, and affirmative) field, an antonym dictionary that includes the (antonym, an accent, a negative score) field, and an antonym that includes the (word, antonym) field. A total score of a word and a total score of a counter word, wherein the antonym dictionary is such that the disagreement intention is not regarded as the intention of consent or the intention of consent is not regarded as the intention of the opposite. .

Images