TW201007482A - Article generating system and method - Google Patents

Abstract

The invention provides an article generating system. The article generating system comprises a general word database, a specific topic word database, a judging module, a determining module, a selecting module, and a combining module. When a specific topic is determined, the judging module generates topic keywords corresponding to the specific topic according to the general word database and the specific topic word database. The determining module determines a plurality of filling words from the specific topic word database according to the topic keywords. The selecting module selects target keywords from the topic keywords and selects candidate filling words from the filling words according to the target keywords. The combining module forms an article related to the specific topic according to the target keywords and the candidate filling words.

Description

201007482 IX. DESCRIPTION OF THE INVENTION: TECHNICAL FIELD OF THE INVENTION The present invention relates to article production, and in particular, to an article generation system and method that automatically generates articles related to a particular topic. [Prior Art] © In recent years, with the continuous development of technology in the field of artificial intelligence, its application in digital learning has been extensive, such as computer question answering system and computer self-operating text scoring system. Since the writing of the article is a highly skilled work, coupled with the increasing demands of modern people for writing efficiency, it is possible to effectively use artificial intelligence to assist users in writing early automatic generation of systems or Research related to writing aid systems has gradually begun to receive attention. In general, most of these writing aid systems use a comprehensive corpus containing a variety of non-social files to test the source of the material. However, the disadvantage of this type of writing aid system is that it uses the material of the comprehensive corpus. The resulting text may not always have a consistent theme. The content of the chapter is likely to be too similar to these materials. Therefore, when using the ^ system to assist it in generating articles related to a specific topic, the actual wood f 3 The articles are often not related to the topic or the statements in the whole article are not fluent. Therefore, the user's needs cannot be met. In addition, the target has become the most popular language in the world. The number of people who want to learn f Chinese is also increasing. However, there is still a lack of a complete and easy-to-use secretarial to effectively provide the towel to write _ help Wei, and pay for the confession and fluency. Chinese article. For Chinese beginners and 201007482, it is really not convenient to say this, this is based on the production of pure articles and methods to solve the above [invention] The scope of the invention is to provide - The article generation system and method of the article ❹ 。 。 。 。 。 。 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章 文章A specific implementation Γί = f contains a comprehensive corpus, a plurality of ί ί ί : 。 。 。 ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ^ The utterance of the essay is light and there is special in this embodiment, when the specific theme is broken, the module will be judged according to the comprehensive corpus and the special library == number r main ☆ ☆ = (2); The gamma sac and the plural are determined according to the second stipulation of the present invention, ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ The corpus stores a plurality of general topic articles. Then, the ^=generating method selects a plural number from a plurality of topic keywords in a plurality of topic keywords according to a plurality of topic keywords. Then, the article generating method selects a plurality of candidate padding strings according to the plurality of target key number of true charging strings. Finally, the article related to the specific topic is formed according to the plurality of target keywords and the plurality of candidate padding strings. The third embodiment is an automatic generation system for a love letter system and an article generation system of the first embodiment. The self-recovering oil theme is selected from a plurality of target surfaces. New ugly combination of mining mode.

The subject keywords of the subject matter have been fixed (the sMct ϋ曰 automatic generation system calculates each topic keyword knife by a modified SPLR sPi ^ method, and has a SPLR score higher than a certain topic keyword. A = target keyword. Then the love keyword series teaches each of the target keywords in the early morning to be automatically ΐ 系 is is ίί increase the number of keywords in the keyword list, the love book reads each topic keyword also It is green; it has to be included in the list of words (4) i = key small and 关键 is related to the key. The key miscellaneous 'Ν can also be determined by the user to have the article generation system and method of the present invention The theme of the St sentence is used as the auxiliary f function, and can automatically generate the character 7 201007482 to improve the efficiency of Chinese writing, and also to further understand the figure. [Embodiment] The present invention provides an article Φ ❹ 可 = The problem can be written and _==, according to the first embodiment of the present invention - the article works on the principle that the first one is related to a specific topic and the key ί 宝 宝 》 Words from a number of specific topic articles 忑 a neighboring key = a number of candidate filler words between words. : Good; ^ [Title ί: 广: 彳 彳 实 = = = = = = = = Many correspond to the balance words provided by different research institutes containing about five million financial words 'for example - : : ; ;; if 14 — to specific meditation (10) to reach 201007482 = two = ^; 2° system _ take · The functions are detailed in the first one. The modules included in the secret 1 and their chapters. 1G saves the city - the general theme text is fixed; ΐ stores the number associated with the - specific topic ίί The problem determines that the specific main super π corresponding to the article to be generated is divided, and the judging module 14 may include 'the soil φ map' for the user to select from the plurality of candidate themes. It can be automatically produced by the article production system i. The special subject is determined. After the specific theme is determined, the judgment module w 1〇 and the special __ 12 recapture library article, the monologue and the carved rod ^^^^ The plural words related to a particular topic are also considered to be word units. 疋, 疋 all punctuation examples and §, hypothesis The user has selected the theme of a report, and the word unit naturally selects a specific topic from the topic as "Class 18 | See the number of candidate clocks displayed by the household 5". The total subject corpus 12 total right = should be "study ten "Minutes" related articles, the syllabus will be combined with the "green slogan and other processing steps, ^ chapter through the "sentence ten minutes 7 " τf 百文早中的复数201007482 then, will introduce Judging how the module 14 produces a number of lines. _ As the name implies (4), it is a word that is less frequently appearing in a particular topic article. Therefore, it is judged in a topical article in a specific topic article. In the middle and the general Hi1 4 ==: word unit, whether the word unit is the subject keyword. : 2: The fd to the unit appears in the frequency of the specific topic article - I right S S language is very low frequency, the word unit has the theme keyword in the 'transfer ^ topic article.卩 疋 疋 Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ Μ If the same-word unit is in the ί frequency row = the first - the difference between the ranking and the second ranking is premature. 〃 1 value' to determine the threshold value of the 5' ranking gap can also be set by the user. Going to pick the gap between the boundaries of the hour, the name of the M-field name will be quite a lot. It is likely that many of the themes of the theme will be turned out to be very rare or not. =, , ^ paid to = the setting of the boundary value is appropriate to the theme keyword of Jing Module 2. This room is successfully passed to the useful example. Suppose the specific topic is “One less than one (8). According to Figure 2 (4) and Figure 2 (9), the word unit “Next n 10 201007482 is ranked second” but in the second ranking No. 6359, this generation p; less than the work of the clock (10): the first -ms of the word; the second "the second is in the lower 氺 氺 扭 扭 二 并 并 并 并 并 并 并 弟 弟 弟 弟 代表 ^ ^ ^ ^ ^ ^ The name gap is much smaller than the threshold value. ϋ - Sort 2 "" is selected as the topic keyword.靖This Jingmo, group η will not be the plural _ female theme:;:; ί theme corpus 12 峨 i: s?ss 主题 theme class 16 will be for each specific topic article , and the word string.岐 ‘) The theme keyword of w will form a key. There are few people, more or more winds. The ten records of the ^^ 201007482 or i: or less have certain use. "." " ί.2 string per t.i by the part: the main: _ word formed by or after the part of the theme keyword in the middle 'when the reliance group 16 has decided to fill the number of padding string ί: ^ Ίί^^ Target key 峨 plural number of filled Xie complex number of display target keywords. In this embodiment, the purpose of (4) a plurality of types of key word string is selected to close ΐ ? 定 主题 主题 主题 主题 主题 主题 主题 主题 主题 主题 主题 主题Number, solid: word=needle Ο Ο ί ί ί 想要 想要 想要 想要 想要 想要 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ After selecting the module 18 to select the string, the selection - 18 will be based on the key string:: middle Ϊ two candidate target words of two adjacent target keywords, to: 12 201007482 select the padding string Between the key words. Appears in the two adjacent objects in a specific topic article: between the previous decision module 16 for each text string between the keywords when the keyword is marked 'As a selection model

G ❹ In practical applications, for the convenience of the user's operation, the step of sorting the candidate fill strings and automatically filling the candidate fill strings of the gates into the two adjacent target key users For the article generation system i to automatically fill in the thinking, the user can select the basin subtraction string from the other candidate padding strings. In addition, if the user is full of thinking about all the candidate 3 filling words, the user can input the appropriate filling by self-input, and select a plurality of target riding and corresponding to each two phase brakes by the above method. After the plurality of candidate padding strings of the keyword, the combination module 2 can form an article related to the specific topic according to the target keywords and the candidate padding string. Next, a detailed description will be given of the operation window and operation flow of the article generation system 1 by means of the diagrams of Fig. 3 (Α) to Fig. 3 (Ε). First, the user will enter the article theme selection window of the article generation system 1, as shown in Figure 2 (Α). In this window, when the user selects the topic of the article to be generated from different article themes (for example, ten minutes after class) and presses the button of 〇κ, 13 201007482 will enter the window of selecting the length of the article, as shown in Figure 3 ( Β) shown. _ In the window of the length of the article, the user can choose to produce an article with a longer, medium or longer length. Since the length of the article is related to the number of keywords it contains, articles of different lengths contain different numbers of keywords, such as s, for example, or three keywords. In this example, suppose the user chooses to generate a shorter article (containing 1G keywords) and presses the button of 〇κ, then it will enter the window of selecting the keyword list, as shown in Figure 3(c). Show. ❹ 于 In the window of selecting keyword list, the system will preset to sort the first keyword list, but the user can also freely select from other groups of keywords listed in the page. Make a choice. When the user completes the action of selecting the keyword list and presses the 0κ button, the user will enter the article editing screen shown in Figure 3 (D). Tian ^! 1 ^ three 0:)), shown in the article to change the face towel, the secret will be based on the keyword selected by the = person! "Short... soil... full double... treasure blame...a +... gue...grind...plate...Golden seat" Automatically produces f-preset text i ±^' very charm. The thing that the seams do is not to go to the temple, but Chu can’t do it. And ride it in the article to edit it. In addition, the top of the article editing screen shows the 10 keywords and the filler words included in the default article. The user is not satisfied with the keywords "minute" and "valuable" in the default article. The user can click on the F3, and the filler candidate window in the article editing screen will be displayed. k & fill in the filler word label F3 and has been sorted by a plurality of candidate fills for the user's face as shown in Figure 3 (8). The first word is the ", very" selected for the default article t. In addition, the candidate fill level is not satisfactory, and the immortal can also be the lion fill word candidate. At this point, an edit window will appear in the page editor 14 201007482 for the user to enter the text as a fill word for the fill word label F3. Xu", "Probably make a choice. ❹ Xin a Shi in the application, when the article generation system 1 provides user candidate fill = can further increase the function of synonyms. For example, if the original pre-article selected The closing word is "Possible". If the user wants to use the "Fff Synonym, the user can select the function of the synonym. At this time, the article, the student system: will list the candidate synonyms of "may", For example, "or "maybe", etc., the user can be among these candidate synonyms: Yu: Rhetoric skills, therefore, the article generation system is for example "like", "seems"; according to the second embodiment of the present invention Flow chart. q Four series of paintings should not be the title of the article: the specific main out. Next, the method performs step Si2, and according to the 15 201007482 inscription corpus, the other subject words corresponding to the mosquito main form are generated. The specific subject corpus stores a number of related topics related to a specific topic. The main part of the syllabary of the syllabary of the syllabus is 复 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于 于The difference in the frequency of occurrence depends on the early ^^^ as the subject keyword. That is to say, the '^--word unit appears frequently in the special topic article and appears low in the general topic article, that is, it means that the word unit report may be the topic keyword. ..., In this embodiment, it is assumed that the frequency of the specific topic corpus two = households in the pepper topic article == :J; the word ϊ = - the ranking has the first - ranking and the (2) * can be selected as the main face _ . And 7〇 whether the second rank is 2434, 2 and is sorted in red, the critical value of 'therefore, the unit "mechanical is the difference of the main: = the part of the word is marked. Then the square has a look at the two keys Borrowing between adjacent keywords: every 16 201007482 g method can capture all the marked padding strings to get the plural number of paddings: 6 'from the plural topic keywords to be used by the user In addition, the plurality of target keywords may be step s, according to steps, and selected. Then, the method executes a plurality of filling side key words, from the ❹ Ο and the two ί the plurality of target keywords The substitute target ^ key selects a plurality of target keywords according to the plurality of target keywords and selects a plurality of (four) generation fill word strings, and sets the main keyword and the plurality of substitute fill strings to form a special string. The specific embodiment is an article of the automatic love generation system ί m ^ This love letter automatic generation system and the specific embodiment ϊ ΐ 词 词 词 词 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 及 词 及 词 词 词Point to explore. (4) ϊ ί ' Regarding the automatic generation system of the love letter, the method of the keyword + keyword is mainly used to calculate the SPLR score of each topic keyword through the improved SPLR side, only when a certain master = When the SPLR score of the key is higher than the critical value, the topic keyword will be selected as the target keyword. own The so-called SPLR method is generally applied to find the unknown word (unkn 17 201007482 word), in order to The spLR method is applied to the love story. After many experiments, it is found that if the method in the original spLR is removed, the best result is obtained. Therefore, the love letter automatically produces ^^' using the modified SPLR green, The age question _ word is 'important and can represent the target keyword of this article' and then these are for the user to select. The improved spLR method can be; lower &

SPLR

Max (tf {KW_L), tfiKW^R) ) 5 KV^~le^ > 1 (1)

m-L

KW;_L KW^len ~2 ~ (2) = (m; - {KWi L |} '一" (3) KW ΐ IΠ production keyword; KWLlen represents the keyword number L represents the keyword The left half of the .p right half; 嶋 嶋 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ In special circumstances, that is, when the article is established, *, and, there are 3 waiting for the meeting; after that, there is no subject key $ early, the SPLR method is evaluated: the system will close all the topics. Therefore, the processing of the love letter ^ (n_lizing) SPLR score as regular 4 will be distributed between 0 and 1. The SPLR score name with the keyword of the subject is in this embodiment, 'after the above - the keyword will get - one is between 〇*! threshold, one master is the main face of two words _ can make 0, the number : Its t 'because the young party has less than the intention of the system, the household, J elbows served 0 points. After that, the love letter has the highest score of five main __ as the target __ since 201007482. The number is taken from the score S: value: after: also:, the SPLR score of the R key word is greater than 〇3 eight _ ' = 〇 Only a certain topic care = the number of i_R knives on the main structure is less than 〇·3 points The subject keyword will not be _ target ❹ key word I: bitter = 3⁄43⁄4 the two SPLR scoring methods selected in the article. θ, the theme as a practical example to illustrate the Lord in the main shed _ "sweet and bitter" example, if the mouth of the rain theme keyword left half, you can get "_i * only fuck the right half of the key, then You can get " ", and you can only take the theme of "Sweet and Sour", "Sweet and Sorrow" and "|" °H. In this article, the keyword "Sweet, Spicy, Spicy and Spicy" appears 10 times, then the subject is sweet. The SPLR score passed by the ancient spicy is _Q=i points. "苜" In the example keyword "Student President", ~to the left half of the main _ key word, can be; ^ U page takes the right half of the key word, you can get: live meeting", right age only theme The "Student President" appeared 2G times, 2 in the article, the President" appeared 20 times, then the main rhyme _ 彳 2, appeared 40 times, "the SPLR score is 2_ =,, key / Shengsheng President The theme keyword "sweet and sweet bitterness is compared with the above two examples." The SPLR score is higher than the "student's goal is _2 and the R value is f G·3. Then you can be "sweet and bitter" can be seen = PsLR2 threshold is set to 0.6 points, no. The system «key. But the "student president" is 19 201007482. In this embodiment, the love letter automatic generation system can select a plurality of subject keywords in each of the plurality of topic keywords included in each article through the SPLR = sub-method described above. Off ^], and these target keywords can be formed into a keyword list. For example, the love letter is self-sincere (four).彳 _ key word string User selection. The main reason is that Wei can hope that the user can directly determine the content of the article he wants to produce from the target of the four-dimensional _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ,

Next, the method of automatically generating the system expansion keyword string will be introduced. For example, assume that the key word string obtained through the above SPLR scoring method includes a total of five target keywords. In order to increase the number of f-keys in the keyword list, the love-writing automatic generation system can respectively regard the N-character keywords of the target keywords in the article as the target key 3 and the order N and the The number of keywords to be included in the keyword series is, and N can also be determined by the user. It is worth noting that the automatic generation system of the love letter adopts the intersection method. ^When the different keywords are selected, the same topic keywords are just taken when the N keywords are taken before and after. Repeated selection of the occurrence of k occurs. Next, the actual operation window interface and operation flow of the love letter automatic generation system will be described in detail by Figure 6 (A) and Figure 6 (B). First, as shown in Figure 6(A), the user can select the length of the love message content to be generated, the required keyword list, and enter the names of the sender and the recipient. When the user presses OK, they will enter the window of Figure 6 (B). As shown in Fig. 6(B), the love letter automatic generation system automatically generates a love letter based on the key 1 string selected by the user. If the user is dissatisfied with the content of the love letter automatically generated by the system, in addition to pressing the randomly generated love letter button to regenerate the love letter, the user may also choose to re-select the keyword, self-innovation, re-election/replacement 20 201007482 ' The original love letter was modified to meet the expectations of users who have functions other than the extension of a word. Compared with the prior art, it is effective to provide a maker's towel according to the present & ^The secret and method of this chapter in addition to the system, therefore, the article has market potential. In addition to the efficiency of financial writing, it is also very

The features and spirit of the present invention are described by the preferred embodiments of the present invention, and are not intended to limit the scope of the invention. On the contrary, the arrangement of the objects and the equality are arranged in the "change and equal arrangement.", and all possible modifications are made. 201007482 [Simplified description of the drawings] Figure 1 shows the third aspect according to the present invention. The functional block diagram of the article generation system of the specific embodiment. Figure 2 (A) and Figure 2 (B) are the first order and the second order in an example respectively. Figure 3 (A) to Figure 3 (E) are respectively The operation visual interface is shown in an example. _ Figure 4 is a flow chart of the edge method. The detailed flowchart of the step S14 in the fourth embodiment of the present invention according to the second embodiment of the present invention is shown in the fourth embodiment. The fourth embodiment is based on the love story of the embodiment of the present invention. All the screen windows of the honest shirt.

[Description of main component symbols] S10~S142: Process Step 1: Article Generation System 12. Specific Subject Corpus 16: Decision Module 20: Combination Module 10: Integrated Corpus 14: Judgment Module 18: Selection Modules F1~F10: Fill word iron 22

Claims (1)

201007482, the scope of application for patents: 1. An article production method, including the following steps: (a) Deciding on a specific topic; 艮--Comprehensive corpus and "more subject matter" to set a plurality of topic keywords, ^f the synthesis The corpus stores a number of general topics ^疋 Bishop early (8) According to the view: he asked difficult words, from the special decision to a plurality of filled strings; suspicion ^ title early self-recovery a topic keyword _ select a plurality of goals Key words. According to the plural target test words, from ', a number of candidate fill strings; and a true charge string. The selected 2 basis, a plurality of target _ words and the complex 2, 3, 4, Refer to 5 to form an article related to this particular topic. a Authentic substring The article library described in the patent application scope is a large balanced corpus. / The comprehensive language ^ Apply for a patent! The article production method described in the item, two items: from: key; now in the specific subject corpus = - H word 'needle' the first difference between the fresh and the second frequency is greater than the article described in the first item Method, the towel step (C) package (1 indicates the plural _ key word in the plurality of specific topic articles; to mark each text segment between each adjacent keyword as a fill string from the waiter to the plurality of The method for generating an article according to claim 1, wherein in the step (d), the plurality of target keywords are used by a user from the plurality of topic keywords. 7. The article generating method according to claim 1, wherein in step (d), the plurality of target keywords are passed through a modified spLR (strkt _ase like imoodratio) scoring method from the plurality of topics. Selected from the keywords. 8. The method for generating an article as described in the patent application scope, wherein in step 1, the user selects a plurality of alternative targets based on the plurality of target keywords. The word ' ^ is based on the plurality of descendant target keywords and the plurality of candidate ‘charge strings to form an article related to the particular topic. , 9, = Shen ^ patent range! The article generating method according to the item, wherein in step 7: the user selects a plurality of alternative sub-strings according to the plurality of candidate filling strings, and forms a pair according to the plurality of target keywords and the plurality of guest strings Articles related to this particular topic. 10, an article production system, comprising: a corpus corpus 'hybrid corpus 赖 有 有 - - - - - - - - - - - - , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , The subject article, the special topic article of the ceremony is related to the specific topic, the complex number is flat and the verb is coupled with the corpus of syllabus _ subject corpus ^ = θ when the specific topic is determined, the full = library and the The specific subject corpus is generated corresponding to a plurality of topic keywords; the genius Wang Zhizhi is connected to the plurality of subject corpora and the judging module, and the subject is reduced to a specific article, and the singer is enchanted, the selection mode, the "multiple charm side key" Word cap ^ plural target 阙 key words, 24 201007482 =^=^=龙狮勉 string cap —=s'ssins='n__ number of topics related to the article selected fill string formed with the specific 仏:27:^^ The article generation system (4) in the comprehensive language 12, ttf profit range, the article generation system described in item 1G, i sentenced to _ 13 if the model is the female patent, the scope of the patent is the first; Question; _ vocabulary: one or two times The number is a first frequency 'the;;;==; ϋ 为 is - the second frequency', the judging module selectively depends on the first word single 14 m according to the first frequency fish 4th - the critical value 15 to ssr A text segment is marked as a -fill string to obtain its, the article generation system described in item 1G, wherein the user of the selected model is from the plurality of subject keys 16. Patent scope 1G An article production system, wherein the selection model mountain 2, a modified SPLR score method selects the plurality of target keywords from the plurality of topic keywords. < 25