1237188 玖、發明說明: 【發明所屬之技術領域】 人類語言邏輯是項立體思維函數反應,故電腦上也應建立至少二個 平面上的思維平面,建立思維平面,簡稱思考平臺,其方法是先將中文 (含其他語言亦同)字或詞,視爲唯一標準的數碼公式定義開始,稱爲 語固基因碼初步。是語言語意分析的基礎以及目吾文付號歸納的切入點, 其與子句句型歸納區域定義和程式流程三種要項是構成使用電腦思考 平面的方法。 【先前技術】 世界歐美等先進國對於中文化電腦以big5或properbility機率方 法;由來有二十多年了,但仍無法突破對中文字等運算技術,如以爲中文 字是一個字一個字型態之視覺誤解,乃至於另種文化用詞如:漢字基 因,乃爲人爲擬人化biology,以其電腦智能仍難以運算、若祇是古文 定義之局限,皆無法應用數位化而瞭解中文語意。 原來傳統的語言語法科學所採用的的頻率數理論,並不適合臨時出 之中文句型,其原因是1個詞就算是百分之九十機率,但一句話有5個 機率相乘後,就祗有百分之四十以下的理解率,且解析語意光是歷史資 料庫詞構機率(Probability)是不夠理解臨時之情境的句型,所以瞭解 一個詞構或子句也需有各種前後文句文化函數符號之相關性才足以定 義句型出語意者。 1237188 【發明内容】 本發明提供-種有關語音及語意符號辨識,其係由多重函數如語法 符號及文法符號所歸納,可以先依片語的組合模組,先得出各種同音異 義詞構的縮小範圍,再依矛盾方程式(即有A就不等於B)原理,將文 法詞意中,共屬語意的屬性如各種文化共識的句構詞構之歸納方法所得 之詞性基因符號,稱敎法制舰,如此提高了語音轉文字之辨識精 確率再與配合語態㈣,更提高了語音辨識能力,形成本言文法模 組,自然就可以對應出另外語言的相同的原理,其所得之文法模組符 唬乂原扣句、片语極大化取值,所形成的文法構詞,所能對應的詞性 符號組合,再由資料表圖,所列之模組序號相互對應,有符合者之後, 就將返回原語句的字詞構序號,如此逐—詳細對應各種原語句所出示過 的字詞構及其雛槪,雜何語言上的雛贱,因此可成為世界語 言一條鞭的格式就得·成;每種平難標表格純都可賴常補充, 依最新語纽新糾顯充至最雜,故在«術語及人士之慣用語也 可適用錢方法將,吾意袼式歸納出數仟種的語意極大化模組序號,可 以使任何—句複雜的句構,拆合成主要句子語纽補充句子或片語等, P使用纟可以代表一詞或一句子語意,此為電腦識別語言模組符號之 σ y亍有放策略法則’故4構、句構乃為有文化數學建構之科學内容及程 式模組意義者。 7 1237188 【實施方式】 〆』㈣構建’為四碼加—個參考碼共為五碼,其為⑽ 尸:為Γ、2料种文雜音之魏音如:校了^4,首音為丁 :二 _ 4作參考。介音如-或如…V…為稱介 其先㈣考慮,㈣3、4格騎财的料代表依人工定義取得 後之發音首音如校為木及交得為门叫,故校字基因碼就成為丁么门q 4等故王辦文子詞乃至其他語文皆可有唯—對應的文字基因碼。每 字也有_碼為首尾音顧,作為語音的分私具,容魅^每字詞 在X抽上另有個詞性符號為ABCD四碼構成,也可在χ轴上另有個英 文原形詞對應因此字構表可有數萬格在γ軸上,各分佈在γ轴與乂軸 對應之函數值上,依不__文喊,在觸方面各種敎依格種一 般詞目或是口語詞及詩詞或_,組成γ軸,每詞也有基因碼,其構 音=部i (第-序碼),第2字為部2,故每—字詞基因碼為五碼構成, 可成為億萬辦碼容量,足夠各糊句唯_對應基因碼,簡碼為首^尾 1百2尾2及〇構成,如地理n_〇另詞性符號為五碼,其定義 原則另述。詞也有各國語言的翻譯格式,其中包括各種文法時態及前後 相關的介詞或碰詞規定。依X軸方向延伸列表,以便電腦依χγ軸對應 得出表格中之合適的字詞翻譯格式或基因碼。詞碼可容有百萬組數量。 語言基因詞性符號是一種文法歸納符號,不僅是先對詞類大項分 為:語態詞、動詞、副詞、名詞等(第一格定義)。及將同函數詞屬性, 如感觀類或同事屬詞性,這些依語態詞相關而構成如很吃驚,很吃緊, 1237188 其%緊與吃驚相_詞性。 的歸納符號排列。 〃子。1的組成可成為短句,或子句 至於三字詞基因碼則可為首i首 定義序列外理不拆、 毛|3尾3 (相同)(可改變 義序亂原理不交)四字詞構為首丨首2首3首4尾4(其語 i五州構可為首1首2首3首4首5 (語音簡碼也相同)六字 九/因碼為百1首2首3首4首末,(末代表末字琍七字詞八字詞 ^詞也可心掏目嘱。但財撕料,變化取位 數’赠答案是電腦中唯一對應的基因碼或基因簡碼就是最基本方法原 貝曰J 了’子構列構為有其各種基因碼定義之字詞數學函數構件者;句構亦 疋_ ’。此表之目的祕:t—㈣現時峨__成數段落 再^之_㈣職曝觸恤峨來。再由詞構 。予.¾的雕基因付號她成與另表對應出概組合或子句構而來。因 此可以得終句詞的語意,此語意也可對應出相同模組之語意符號的不 同國語文句型且其時!!等皆可依代號設定值符合正確合理文法轉換規 格。 這是表格靖輯法觸餘合絲式的優點,_祕方程式的文 法格式。 例如:“架設在地平面上的—種鐵架稱域架”;首先依推碰合三步 驟組成的基因碼’推理對應出字詞,得出架設/在/地平面/上/的/ 一種/鐵架/稱為//鷹架。比照資料内容至詞性符號的組成可對應出句 構之詞性組成=V2600 + W3000+N2600 + B3720 + W2000 + NMG20。 1237188 其中有如:NMG20 = (MG200) + (N2600)詞性組合,与子句詞性組成= (V10〇〇) + (N2600)所以本句話可合為二個子句構成’(另。將字詞含不 同種文字單字形式)依其形式或音式形音合併,及筆劃數及文法詞性基 因符號等作為八碼同模組格式,此詞構依基本公式,如二字詞為首^、 首2尾1 、尾2、法卜法2醜共為八碼,此八碼組合數有6〇 的八次方’約有數拾億個詞構容量,此可容下世界各國的詞構總量及總 合,且是唯-對應的基是符號語意分析的初步手段,上述劃數, 是指筆劃數的總和,可贿得,不管任何增加之_之_先後及齡 ^ 之不同的碎師定義,祇要制此五或㈣組碼公式,就會有一樣的基 因碼型式,並產生唯一對應的詞構基因碼。 用此組媽就可形姐界語言轉數碼平臺,(如圖_所示)其為簡 化之數碼内容;此詞構平臺同—個水平χ轴位置,不同欄位上可表示 ^另外同-文字的函數對應詞構,如文字(_)(二)故簡、繁、日、 韓、越、英、德、法均可直接對譯詞典或相異的文法詞性,制三字詞 (單字形幻數碼編號也是同理,四字、五字、六字至十二字詞都㈣ 模組詞構基因碼(皆由單字八碼中抽出形、音、形音、法、書⑻且如 輯於快速對躺歡義甚钱助,可省鄕日植。另也可含5字6 字7字詞的詞構稱紋義之語態詞區,皆可自行自由絲之,其為了縮 短句子長度及判讀更簡化之目的,即首先在一句話中,先找出語態副詞 作切割句型分段者(或稱碰詞、或稱踫字),因其更容易快速區分每一 · 句話為2、3段子句’及較少詞數目組合數目群,而可快速正物合出 * 10 1237188 同資料庫基因碼對應詞構,如有相接處有同屬左右詞構之跨字成為各可 :之觸時’可依以下三段原則處理⑴是内搶:即為某字二全在二 字詞之間存在,此由字數多者為主詞構決策,少數字詞者即被併合⑹ 是外搶:某字詞—部份係與別_鄰字詞—部份相蚊,其決定搶字的 方法是以靠語態詞的詞構為優先,如(應該)建立正確姿勢,,構‘‘建 立”比“立正”織“翁,(語陳―)考“以為-詞 構’本原則亦含語音簡碼之定義,如語音符號:的时,是尸尸,在p 巧’等是為了段落句«為數段短辭句,錢少_構組合數目成為 語音式語意的語意首尾波形對應,來對應表二的基因碼對應χγ抽内 容,成為符合銳碼龍庫之觸和子句,觸_彙之數學模型(因 代表另個對應之基因碼内容)。 另外搶字的處理可用某句話每字(如圖三)為首排,然後接著組合 本句詞構,(在12字詞之前均可組合對照圖一表的基因碼),再看某字 框有無與前(後)字排框有搶字(内搶不計)其判讀流程如圖四所示。 但不得違反例外之法則,即在某些指定的詞構中是依前或後的詞構 之文法符號定義決策(圖2)而作的拆解或合併的定義,如 字詞:如 前文法相關符號 後文法相關符號 _拼合結果 果然 NO WA _果+然 果然 WA+NO~~一 NO 果然 1237188 故一句中“果然”會依前後文法符號而變為果+然(分開)或果然 (合併的精確詞構。此例外法則係在字排表(圖三)成立後,即檢查有 無某些字詞構先行拆合而不用組碼其與語態詞相同意義的分段落而分 別計算詞構的基本作用。以上如此產生了這句話的基元單位,簡稱為基 元排序。 基元排序為本句話的詞構單位及其相對應的文法符號,如:喝了水 果然來到樹下。基元=喝+ 了 +水+果然+來+到+樹+下本句詞性符 號= VO + WA + NO + WP + VO + NO + BD。這是分析語意的基本階 段,以圖四方法加上語態詞的方法就可以使一句中文(或其他文句)產 生合宜的基元,故每一字詞皆有相對應的函數值如圖五所示,其丫軸為 各字詞,X軸為某詞的文法符號,指標以填充符號〇p英文或德文文法 格式轉譯内容,或其他可應㈣哲學代號,使成為人輯某詞的意義皆 可錄下,故Y軸的各種字詞,就可以使得圖五完成某種語言文字的基本 邏輯及文法格式平台,而填充符號0P代表著物理函數值之填入者。 再將如中文的片語(如圖六)作為判讀句型文法的最小單位或最需 先組合者’如數詞1,2, 3 (GZ符號=MM),樓(GZ符號二⑹相加 (MM + GG=MG)或吃了,=VO + WA=v〇=^GZ,將新 Gz 填入 GZ攔中(此為資料庫之一),因此再進行到圖七的跨片語組合表,内填 可組合之觸式但料同觸所能構成的“句中各觀構觸,,如美麗 +女孩= JO + AM,地球+賴層=M+M等等,如此增加片語的組合 特性以及圖八為片語子句的組合表,如:我的_種發明=我+的+一種 12 1237188 發明=NP + WA + N〇=NO (資料庫之一);再來組合子句(如圖九所 示)稱為語態子句組合表:如將σ語制語,_帶補文法符號可以為 成慣用子句者如:若不是天氣的緣故=若不是+天氣+的緣故=ra+ NO + WA+NO=RA+NO填入表格中,使成為語意模組單位之一,而 且有新的文法符蚊義’但此皆敎法符餘_代敎字詞,所以一 侧模組基關單位可替代千·文字贿量,十倾組就可以有千萬 個-般同雛字詞組合,又加上其他符號如“〇p填充符號,,作為另種 函數值定義填此0P者料’就足夠以少量的ID數目排職表億萬個 人類會用的顧語或特咖語的函數f料庫細,再加上圖九的句型定 義表,其t GZS極大模蝴巾文模崎某國語言㈣的基本格式,用 前述的圖五、m ·、圖九各模崎號所演算組合或再組合 的符號’構成-合法句型崎應過程,使得電腦資料庫上有完整的語意 符號總體(包括句、子句、片語子句、片語、字詞)結構,而且將臨時 之句型係由這些文法符號極大化(最佳化)對應資料庫後組合而成(圖 九所示)’所謂極大化係指能成為句子者,就不以子句出現,或能成為 子句者就不以片語出現’但若有其中有外搶現象時可_十—流程方法 解決搶片語之繼方式’如依序由小而纽合後再將前述片語子句規定 再組合-次(因為組合句的文法符號,可以再組合新的文法符號)所以 再來回組合—次,並且㈣顿_特定子句文法舰絲能成為主句 的片語文法符號及含標點符號的小句文法符號暫時移出,另作極大化處 理’以後再合併成柄構’此稱邊除化鱗除化;使之謂應到極大化 13 1237188 的本體㈣制。料種顯序麟《财«红讀尋句構模电 序叙各種問答句型的群組序號,以指定之語態詞構及加分比率,毅 案内容之語意序號相對應作成總分比率,即成為—種用問題對應知婦 唬mdex後,再用問題模組序號线態詞構符號,句構中尚有—等除化 之片語式子句,係在定義區中可先獨立存在的,如以語態詞為首之子句 或不與句射駐句杨_相併合的子句雜以語魏稱為邊除 化子句。 ^如組合不成的片語或是相等片語可以到最後的句構(句構〉句子) 集合’使件人類-句話最後變成數個字元符號的集合,此就是人類中文 翻譯成電腦中文語意符號的步驟,故運用有邏輯符號(包括文法符號或 各種0P填充文法符號)的平面運算之建構一層或多層,相互演算合併 如圖十-所示;構成_思考立體邏輯演算法,可以將人類立體邏輯的語 思瞭解用電腦語言基因(文法符號)資料庫做成的多層次平臺基因運 异;其所擬的基因體化符號而且可以愈來愈豐富及精確及即時以網際網 路的自動填補更正策略,使得此語意模組平臺可提供給每個網友或公司 使用瞭解知識句型之語意者。以上若有子句細子有搶片語時 ,可以用 …Μ為優先取值方法理念作成優先順序,且如某個崎—句話語音資 料其所對應的同模組(圖八、圖九、圖十)若與標準資料庫有差異時, (此為$有的機率)如在人類可以聽的懂的,但在電腦中若用模糊比 對’則會失誤連連,故此可用ΑΡΟ模組技術來依精確歸類將基因符號 先合成模組’如圖十二所示,例子是··有東西可以吃嗎,或:有什麼東 14 1237188 西可以作飯?或例如··你好 好?/皆可符合圖十-之爸爸的同事可好?/老師最近好不 ,h 十―之—的各種詞性模組符號項人娜填充模組等 等,^是辨識符號非僅辨識文子或語音之觸填充方法,如相同語意 … 狀札不同鱗的句子會太s,貞1m電腦觸apo表中 去可以將其基疋符號對應其標準句型示各單位模組之符號 上,在每單位模組符號上有+及±記號,如是+者表示必要有的文法符 =目同者’获堵絲可贿或沒有敎法舰_者,若是像時間 田m刪或地方副詞(伽)則可以列為標準資料庫句子符號意義 的附加及另外比對資料者,此在句型文法上可以稱為邊除化或等除化技 術’此因不影響主要句子語意的結構。故可以將公司内各有關人事時地 物的詞構短句建在資料庫中,及至電腦回答用的英日文為回覆句型來 電問者的詞_及語態詞都是可㈣或對應的糊續於配合之語 態詞句型’用-般祕填充絲式,構成跳軸合魏陳術手段,如 有些詞構子句還可以在一種為〔括弧〕的指令架構下,可對調符合句型 中詞構或短句的基因符號如下所示·· 〔我要買三鮮電股票在5〇·5讀辦〕其可轉於〔請賣華電 股票在50.5元時賣三張〕都是在基因碼先決定後再用詞構或子句的 模組化詞錄因符號’在歡減内’且可對雛置及檢麵件等功能。 因此對於某句型(圖八、圖九、圖十)或片語子句就可用一個Ap〇 參數表APO是資料庫中填充符號比對值格内符號,也稱細模組符 號,依最佳化數學率比冊號其參數烟定義者,視糊_句型的id 1237188 號瑪’而此ID號碼(圖十二)和標準同模組(圖八、九、十)①號是 相同定義與翻譯的’經實驗證明社方法確實可行,且符合人類語言語 意辨識行為,如此可以解決相同語意不同字詞表達法的基本問題。林 方法可不限發聲者為中/英/德等語音發出,皆㈣—般語音分碼系統 得到每個字與每财的跡娜其糾了音元素,啸,每字的首尾 曰碼先將句子之巾的—組音碼組合成為翻簡碼,而得㈣應表中詞或 片語之基因簡碼,而找出其語意符號,其之後的詞性符號功能也如同前 述内容者,加上近音組合及錯音組合,近音表示發音者將卜虫,么1237188 发明 Description of the invention: [Technical field to which the invention belongs] Human language logic is a three-dimensional thinking function response, so a computer should also establish at least two planes of thinking planes, and establish a plane of thinking, referred to as a thinking platform. The method is to first The Chinese word (or other languages) is regarded as the only standard definition of digital formula. It is the basis of semantic analysis of language and the entry point for induction of Mu Wuwen's payment number. Its three major items, including the area definition and clause flow of clause pattern induction, are the methods of using computer thinking plane. [Previous technology] Advanced countries such as Europe, the United States, and the United States have adopted the big5 or properbility method for Chinese culture computers; it has been more than two decades since, but it has not been able to break through computing techniques such as Chinese characters, such as thinking that Chinese characters are one character and one character. The visual misunderstanding and even other cultural terms such as: Chinese character genes are anthropomorphic biology, and its computer intelligence is still difficult to calculate. If it is only limited by the definition of ancient Chinese, it is impossible to apply digitization to understand Chinese semantics. The original theory of frequency numbers used in traditional language grammar science is not suitable for temporary Chinese sentence patterns. The reason is that a word has a 90% chance, but after a sentence has 5 chances,祗 It has an understanding rate of less than 40%, and the analysis of semantic meaning is only a sentence pattern of the historical database word structure probability (Probability) is not enough to understand the temporary situation, so to understand a word structure or clause also need a variety of contexts The relevance of cultural function symbols is enough to define the semantics of sentence patterns. 1237188 [Summary of the invention] The present invention provides a kind of speech and semantic symbol recognition, which is summarized by multiple functions such as grammatical symbols and grammatical symbols. Narrow the scope, and according to the principle of the contradiction equation (that is, A is not equal to B), the grammatical meaning, the common semantic properties, such as the syntactic construction of various cultural consensus, are used to induce the part-of-speech gene symbols, which is called the legal system. In this way, the accuracy of speech-to-text recognition is improved, and in conjunction with the intonation, the speech recognition ability is further improved, and the grammatical module of this language is formed, which naturally can correspond to the same principle of other languages. The groupings blunt the original sentence and phrase to maximize the value. The grammatical word formation and the corresponding parts of speech symbols can be formed, and then the data table map, the module numbers listed correspond to each other, after there is a match, It will return the serial number of the original sentence, so that it will correspond in detail to the various word formations and their babies, as well as the idiots in the language. The format of a single whip in the world language can be achieved; each type of standard bidding form can be added frequently, and it can be updated to the most complicated according to the latest language. Therefore, «terms and people's idioms can also be used for money methods. Summarizing, I will summarize a number of semantic maximization module serial numbers, which can make any complex sentence structure into the main sentence, language, supplementary sentences or phrases, etc., P can be used to represent a word or A sentence semantics, this is σ y of computer recognition language module symbol has a strategy of putting strategy 'Therefore 4 structure, sentence structure is the scientific content of cultural and mathematical construction and the meaning of program modules. 7 1237188 [Embodiment] 〆'㈣ build 'is four yards plus a reference code for a total of five yards, which is ⑽ Corpse: Wei Yin, which is Γ, 2 kinds of murmurs such as: ^ 4, the first sound is D: Two_4 for reference. Prepositions such as-or such as ... V ... are called first considerations, and the materials representing the 3 and 4 grids represent the first pronunciation of the pronunciation after the artificial definition is obtained. The code becomes Ding Momen's q 4 and other Wang Ziwen's words and even other languages can have unique-corresponding text gene code. Each word also has the _ code as the first and last sounds. As a separate feature of the voice, Rongmei ^ Each word has another part-of-speech symbol on the X sample, which is composed of ABCD four codes. It can also correspond to the original English word on the χ axis. Therefore, the word structure table can have tens of thousands of cells on the γ axis, each distributed on the function value corresponding to the γ axis and the 乂 axis. And poetry or _, forming the γ axis, each word also has a gene code, its articulation = part i (the first-order code), the second word is the part 2, so the gene code for each word is composed of five codes, which can become billions The capacity of the code is enough for each sentence to be only _ corresponding to the gene code, and the short code is composed of the first ^, the last 1,2, the last 2 and 0. For example, the geographical part n_〇 is another five-symbol, and its definition principle is described separately. Words are also translated in various languages, including various grammatical tenses and related prepositions or touches. Extend the list in the X-axis direction so that the computer can correspond to the χγ axis to get the appropriate word translation format or gene code in the table. Word codes can hold millions of groups. The linguistic gene part-of-speech symbol is a grammatical inductive symbol. It is not only a classification of large parts of speech: voice, verb, adverb, noun, etc. (the first case definition). And it will be related to the attribute of the function word, such as sensory or colleague, which is related to the modal word and constitutes very surprised, very tight, 1237188 whose% tightness is astonishing _ part of speech. Permutations of induction symbols. 〃child. The composition of 1 can be a short sentence, or the three-word gene code can be the first i definition of the sequence without dismantling, Mao | 3 tails 3 (same) (can change the meaning of the order and disorder principle do not intersect) four words The structure is the first 丨 the first 2 the first 3 the first 4 the tail 4 (the language i Wuzhou structure can be the first 1 the first 2 the 3 the first 4 the first 5 5 (voice shortcodes are the same) six characters nine / the code is one hundred one 2 2 3 4 first and last, (the last represents the last character, seven characters, eight characters, ^ words can also be ordered. But the wealth is torn, and the number of digits is changed. The gift answer is the only corresponding gene code or gene code in the computer. The most basic method, Yuan Beiyue, said that the substructure is a component of mathematical functions of words with its various gene code definitions; the sentence structure is also __. The purpose of this table is: t—㈣present 埃 __ 成 数 段^ 之 _㈣ 职 Exposure to E-shirt. Then it is composed of words. Yu.¾'s carved gene pays her to form an approximate combination or clause structure corresponding to another table. Therefore, you can get the semantic meaning of the final sentence, This semantic meaning can also correspond to different national language sentence patterns of the semantic symbols of the same module and at that time !! etc. can be set according to the code to meet the correct and reasonable grammatical conversion specifications. This is the table Jingji The advantages of the method of synaptic synapses, grammatical format of mysterious equations. For example: "a kind of iron frame called a field frame that is set on the ground plane"; firstly, the three-step genetic code 'inference' corresponds to the word Words can be erected / on / ground plane / on / of / a kind of / iron frame / referred to as // eagle frame. Comparing the composition of the material to the composition of the part-of-speech symbol can correspond to the part-of-speech composition of the sentence = V2600 + W3000 + N2600 + B3720 + W2000 + NMG20. 1237188 Among them: NMG20 = (MG200) + (N2600) part-of-speech combination with clause part-of-speech = (V10〇〇) + (N2600) So this sentence can be combined into two clauses' ( In addition, the word contains different types of words and single-word forms) according to its form or phonetic combination, and the number of strokes and grammatical linguistic gene symbols are used as the eight-character module format. The word structure is based on basic formulas, such as two-word words. The first ^, the first 2 the tail 1, the last 2, the Fabufa 2 ugly are eight yards in total, the number of these eight yards has a 60th power of eight, 'about hundreds of billions of words structure capacity, this can accommodate the world's countries The total number of words and their composition, and the only-corresponding basis is the preliminary means of symbolic semantic analysis. It refers to the sum of the number of strokes. It can be obtained regardless of any increase in the __ _ sequence and age ^. Different division division definitions, as long as the five or ㈣ group code formula is made, there will be the same gene code pattern and generate The only corresponding morphological gene code. With this group of mothers, you can transform the language of the sister world into a digital platform (as shown in Figure _), which is simplified digital content; this morphological platform has the same horizontal x-axis position and different columns. The position can indicate ^ and the corresponding word structure of the same-literal function, such as the text (_) (two), the simple, complex, Japanese, Korean, Vietnamese, English, German, and French can directly translate the dictionary or different grammars Part-of-speech, system of three-character words (single-character magic digital numbering is the same, four-, five-, six- to twelve-word words are all ㈣ modular word structure gene code (all extracted from the single-word eight-character form, sound, form The sound, law, and book can also help you to lie down quickly and save money. In addition, it can also contain 5 words, 6 words, and 7 words of word structure, which can be used freely. In order to shorten the length of the sentence and simplify the interpretation, it first finds in a sentence. Voice adverbs are used to segment sentence patterns (also known as bump words or slang characters), because it is easier to quickly distinguish each · sentence as 2, 3 paragraph clauses' and a small number of combinations of numbers, And it can quickly synthesize * 10 1237188 The corresponding word structure of the same database gene code. If there are crosswords with the same left and right word structure at the junction, it can be different: Touching time can be processed according to the following three principles. Yes Internal grabbing: that is, the existence of a whole word between two characters. This decision is made by those with more words as the main word, and those with fewer digits are merged together. External grabbing: Some word—partial and other _Neighbor words—Some mosquitoes, whose method of determining word grabbing is based on the word structure of the modal words, such as (should) establish the correct posture. (Yu Chen-) The "principle-word structure" test also includes the definition of phonetic shortcodes, such as the phonetic symbol: when it is a corpse, in p 'Etc is for the paragraph sentence «is a few short phrases, the number of Qian Shao_combination becomes the semantic head-to-tail waveform correspondence of phonetic semantics, which corresponds to the genetic code of Table 2 corresponding to the χγ extraction content, and becomes the toucher of the sharp code dragon library. Sentence, touching _hui's mathematical model (because it represents another corresponding genetic code content). In addition, the word grabbing process can use each sentence of a sentence (see Figure 3) as the first row, and then combine the word structure of this sentence, (in 12 Before the words, you can combine the gene codes in the table in Figure 1), and then look at the presence or absence of a character frame (before or after), and the interpretation process is shown in Figure 4. But exceptions must not be violated. The rule is the definition of dismantling or merging according to the grammatical symbol definition decision (Figure 2) of the preceding or following morphological structure in some specified word formations, such as words: as in the previous grammar-related symbols after the grammar Related Symbols_Combined Results Sure enough WA WA_Surely Sure WA + NO ~~ 一 NO Sure enough 1237188 So in the sentence “Sure enough” will change to Sure + separate (separate) or Sure enough (accurate word formation combined) . The exception is in the typographical table ( C) After the establishment, it is to check whether some word formations are disassembled first, and to calculate the basic role of word formations separately without grouping the paragraphs with the same meaning as the mood words. The above has generated the primitive unit of this sentence. , Abbreviated as primitive sort. Primitive sort is the grammatical unit of the sentence and its corresponding grammatical symbols, such as: drink the fruit and then come to the tree. Primitive = drink + 了 + 水 + 然 然 + 来 + 到+ Tree + next part of speech = VO + WA + NO + WP + VO + NO + BD. This is the basic stage of semantic analysis. The method of Figure 4 and the use of a voice word can make a Chinese sentence (or other (Sentences) to generate suitable primitives, so each word has a corresponding function value as shown in Figure 5, the Y axis is each word, the X axis is the grammatical symbol of a word, and the indicator is filled with the symbol 〇p English Or translate the content in German grammar format, or other applicable philosophical codes, so that the meaning of a word can be recorded, so the various words on the Y axis can make Figure 5 complete the basic logic of a language And grammatical format platform, and the padding symbol 0P represents the physical function value Fill person. Then use phrases such as Chinese (as shown in Figure 6) as the minimum unit to read the sentence grammar or the ones that need to be combined first, such as the numerals 1, 2, 3 (GZ symbol = MM), and the floor (GZ symbol two ⑹ added together ( MM + GG = MG) or eat, = VO + WA = v〇 = ^ GZ, fill in the new Gz into the GZ block (this is one of the database), so go to the cross-phrase combination table in Figure 7 , Fill in the combinable touches, but the material can be composed of the different structures in the sentence, such as beauty + girl = JO + AM, earth + Lai layer = M + M and so on, so increase the phrase Combination characteristics and Figure 8 are the combination table of phrase clauses, such as: my_kind of invention = 我 +++ 12 1237188 invention = NP + WA + N〇 = NO (one of the database); Sentences (as shown in Figure 9) are called the modal clause combination table: if the σ language is used, _ with supplemental grammar symbols can be used as a conventional clause, such as: if it is not for the sake of the weather = if it is not + weather + Sake = ra + NO + WA + NO = RA + NO fill in the form, making it one of the semantic module units, and there is a new grammatical runemic mosquito meaning 'but this is not the same as the rune word _generational words, so one Side module base units can replace thousands of The amount of text bribes can be combined with ten million groups in the same way, plus other symbols such as "〇p filling symbol, as another function value definition to fill this 0P material is enough to a small amount The number of IDs is ranked by the function f database of Gu language or special coffee language used by hundreds of millions of people, plus the sentence pattern definition table in Figure 9. Its t GZS is very large. The basic format of ㈣ uses the symbols' composed or recombined by the calculations or recombinations of the various model sakis shown in Figures 5, m, and 9 above to form a legal sentence pattern saki response process, so that the computer database has a complete set of semantic symbols ( Including sentence, clause, phrase clause, phrase, word) structure, and the temporary sentence pattern is formed by maximizing (optimizing) the corresponding database of these grammatical symbols (see Figure 9) 'The so-called maximization refers to those who can become sentences and do not appear as clauses, or those who can become clauses do not appear as phrases.' But if there is a phenomenon of outside grabbing, it can be done. "Succession method", such as sequentially from small to articulate, and then regrouping the aforementioned clauses -Times (because the grammatical symbols of the combined sentence can be combined with new grammatical symbols), so they are combined back and forth-and the _ specific clause grammar ship can become the phrase grammatical symbol of the main sentence and small punctuation marks. The grammatical symbols are temporarily removed, and another maximization process is performed, which will be merged into a handle structure later. This is called edge descaling and descaling; it is called the ontology control of maximizing 13 1237188. The material is the obvious order Lin "Cai« The red-sentence-seeking sentence model is used to sequentially describe the group numbers of various question-and-answer patterns. The specified number of morphology and bonus points are used to form the total score ratio corresponding to the semantic number of the content of the case. After knowing the woman's mdex, then use the serial number of the module of the problem to construct the symbol. The sentence structure still has-equal division of the phrase clause, which can exist independently in the definition area, such as using a voice word. The headed clause or the clause that does not merge with the sentence shooter sentence Yang_ is called the edge elimination clause in Wei Wei. ^ If the uncombined phrase or equivalent phrase can go to the final sentence structure (sentence structure> sentence), the set 'makes a human-the sentence finally becomes a collection of several character symbols, this is the translation of human Chinese into computer Chinese The steps of semantic symbols, so use plane operations with logical symbols (including grammatical symbols or various 0P-filled grammatical symbols) to construct one or more layers, and calculate and merge each other as shown in Figure 10; The understanding of human stereological logic understands the multi-level platform genetic differences made by the computer language gene (grammar symbol) database; the proposed geneticized symbols can be more and more abundant, accurate and real-time. The automatic fill-in correction strategy makes this semantic module platform available to every netizen or company who uses the semantic meaning of knowledge sentence patterns. If there are clauses and phrases in the above, you can use ... M as the priority value method concept to make a priority order, and if a certain Qi-sentence speech data corresponds to the same module (Figures 8 and 9) (Figure 10) If there is a difference with the standard database, (this is the probability of $), if it can be understood by humans, but if you use fuzzy comparison in the computer, it will make mistakes, so you can use APPO module. Technology to accurately classify gene symbols into modules according to precise classification. 'As shown in Figure 12, the example is ... Is there anything to eat, or: What can be cooked for the East 14 1237188? Or for example ... how are you / Can all meet the friend of Figure 10-Dad? / Teacher recently, various types of part-of-speech module symbols, such as h ten-to-, are filled with modules, etc., ^ is a method of identifying symbols, not just identifying text or speech, such as the same semantics ... sentences with different scales Will be too s, Zhen 1m computer touch the apo table, you can put its base symbol corresponding to its standard sentence type to show the symbol of each unit module, + and + signs on each unit module symbol, if it is + means necessary Some grammatical characters = "Mu Tongzhe" who are blocked or bribery-free, if it is like Tiantian m delete or local adverbs (gamma), it can be listed as an additional and additional comparison of the symbolic meaning of the sentence in the standard database. Sources, this can be called marginalization or equalization in syntactic grammar. This reason does not affect the semantic structure of the main sentence. Therefore, the company's short sentences about personnel and features can be built in the database, and the English and Japanese used by the computer to reply are the sentences of the caller. The words _ and the voice are all admissible or corresponding. Continuing with the coordinated voices, the sentence pattern 'fills the silk pattern with-like secrets, which constitutes the jumping axis and the Wei Chen technique. For example, some word-formation clauses can also be used to intersect the coincident sentences under a command structure of [brackets]. The gene symbol of the word structure or short sentence in the type is as follows ... [I want to buy San Xiandian shares at 50 · 5 to read] It can be transferred to [Please sell Huadian shares at 50.5 yuan to sell three pieces] The genetic code is determined first, and then the modularized vocabulary of the word formation or clause is used because of the symbol 'in the sorrow', and it can be used to set the baby and check the face. Therefore, for a certain sentence pattern (Figures 8, 9, and 10) or phrase clauses, an Ap0 parameter table can be used. APO is a symbol in the database for filling symbol comparison values in the database. It is also called a thin module symbol. The optimized mathematical rate is better than the volume number of the parameter smoke definer, depending on the _sentence pattern id 1237188 No. Ma, and this ID number (Figure 12) and the standard same module (Figures 8, 9, 10) ① number is the same The 'definition and translation' experiment proves that the social method is indeed feasible and consistent with the semantic recognition behavior of human language. In this way, the basic problems of expressions of different words with the same meaning can be solved. The Lin method can be used to send Chinese, English, German and other voices without limitation. The general voice code system obtains the traces of each character and each wealth. The corrected elements are wailed, and the beginning and end of each character are coded first. The combination of phonetic code and sentence code of the sentence scarf becomes a simplified code, which can be used to find the genetic code of the word or phrase in the table, and find its semantic symbol. The subsequent part-of-speech symbol functions are the same as those mentioned above, plus Near sound combination and wrong sound combination.
今’音不準相混淆則可組成合理的詞構基因碼而用來對應資料庫中相符 合詞性符號的指定詞構來,錯音組合也是 如此,如:太平也洋,其124組合及12·3·組合可以找到有或沒有 之正確雜、基目碼。此是—觀人化思維賴,_隨力分解能力, 故總稱為语言基因資料庫。在電腦硬碟之中或記憶體内運算得出唯一值 出示。Confusion of today's inaccurate sounds can form a reasonable morphological gene code and correspond to the specified morphology of the corresponding part-of-speech symbols in the database. The same is true for the wrong pronunciation combinations, such as Taiping Yeyang, whose 124 combinations and 12 · 3 · Combination can find the correct miscellaneous, base code with or without. This is-the humanistic thinking depends on the ability to decompose with force, so it is generally called the language gene database. The unique value is calculated on the computer hard disk or in memory. Present.
以上所需的電腦記憶體容量大約在6M左右(二種語言對應而言) 運算CPU約在500K内,運算速度可控制在i秒之内完成組合演算,故 實施條件在個人電腦上或PDA上可以完成構建,有了語意模組ID之後 就可以依各種功能用途之平臺應用了。如中文轉英文、或日文轉*** 文,皆是用基元及句子、子句模組符號依不同位圃序數定義出對應之各 種語文的文法格式如aNNi+bPAi + cNi^ + dWA!(中文)轉為英文為 aoo + cNS + doo的函數對應關係,此為一般熟習二種語言文字者,皆可 16 1237188 凡成叹定序數方法之。且關文法符號歸納,故此是全語意句型包括 者,右用在訂講機票或銀行業務諮詢,可用每句話基元對應的文法符號 (稱填充付號)填入至表格内(如圖十三),再將沒有說完或不清楚的 地方用Y軸上AQ獅各個相關用語的填空内標示⑻以解決某些錯 子或錯辨識得之子者,’再以冑綱來訪者,以狀詞意模組化確定後, 將其確定的字詞的基因碼,與同模組基因碼單位(填充模組)之答案計 算區(圖十四表)相對應符合對應,精準且唯一之GW排列欄之項目, 、、二、、心刀數a十异最佳化後就可以輸出,以電腦語言語音(tts)平臺對客 戶回答問題了。如此方式可以應用在各種商業諮詢或訂務上,使得 中文貧訊化技術達到擬人化思_赌,是摘科技的解決策略,本發 明由子的基因符號設立開始到句型語法模組符號的ID排序,以及相配 0〜考机私的夕層平面運异結構,構成有邏輯功能取向的思考計算平 臺’可以解決人關答格式及語意轉換的基本工具;再將此瞭解語意的 方法應用在語言語音的職率,其哲财法是關翻之_文法函數 關係應有存在,朗AP0格式以模組化跳填充符號取得語態句型的語 意符號比對,因此祇要將詞構符號比對套入此語意句型的填空比對即 可,可以處理高誠的字崎紋之拉钱不_切觸構的分析技術, 而可以擬人化的耳力*意邏輯功能、此是種新的配方技術,能確實提高 語音技術喊略之發明,灿TTS直接縣音問或回答問題,每一問 題依圖十三建立之某項專制答對絲,甚至事先將搜尋_網頁知識 作-問答模組對應總表,使得任何網友都可以問到某專業區域知識的精 Ϊ237188 準答案,以操作白板管理語意平台其係—可接受雙方或多方之語言文字 轉對方語言者及同為同-俱樂署理權之社群語意管_統;此發明符 合知識經濟的要件,故為知識經濟之基本工具。The computer memory required above is about 6M (corresponding to two languages). The computing CPU is about 500K. The computing speed can be controlled within i seconds to complete the combined calculation. Therefore, the implementation conditions are on a personal computer or a PDA. The construction can be completed, and the semantic module ID can be used according to the platform of various functional purposes. For example, Chinese to English, or Japanese to Arabic, the grammatical format of each language is defined by primitives, sentences, and clause module symbols according to different bit ordinal numbers, such as aNNi + bPAi + cNi ^ + dWA! (Chinese ) To English for aoo + cNS + doo function correspondence, this is generally familiar with two languages, can be 16 1237188 Fan Cheng ordinal number method. In addition, the relevant grammar symbols are summarized, so it is a full semantic sentence pattern. The right is used for booking air tickets or banking consulting. The grammatical symbols corresponding to the primitives of each sentence (called filling payment numbers) are filled into the form (as shown in the figure). (13) Mark the unfilled or unclear areas with the relevant words of AQ lion on the Y-axis in the blanks to solve some of the wrong sons or wrongly identified sons, and then use the "Gun Gang visitors" to After the adverbial modularization is determined, the gene code of the word determined by it corresponds to the answer calculation area (Figure 14) of the same module gene code unit (filled module), which is accurate and unique. The items in the GW arrangement column can be output after optimizing the number of a, b, and b. The number of heart knives a can be output, and answer questions to customers using the computer language and speech (tts) platform. In this way, it can be applied to various business consultations or subscriptions, so that the Chinese deafening technology can achieve anthropomorphic thinking and gambling. It is a solution strategy for science and technology. The invention starts from the establishment of the genetic symbol of the child to the ID of the sentence pattern grammar module symbol. Sorting, and matching planes of different levels from 0 to test machines, constitute a logical computing-oriented thinking and computing platform that can solve the basic tools of human answer format and semantic conversion; then apply this method of understanding semantics to language The rate of speech, its philosophical method is the key to the _ grammatical function relationship should exist, Lang AP0 format uses modular jump filling symbols to obtain the semantic symbol comparison of the voice sentence pattern, so as long as the word formation symbols are compared It is enough to fill in the blank sentence comparison of this semantic sentence pattern, and it can handle the analysis technology of Gao Cheng's Ziqi pattern. It can be anthropomorphic ears * logical function, this is a new formula Technology, which can actually improve the invention of voice technology. Chan TTS directly asks or answers questions. Each question is answered in accordance with an authoritarian rule established in Figure 13. Recognition-Q & A Module Correspondence Table, so that any netizen can ask for the best answers of a certain area of knowledge, to operate the whiteboard management semantic platform, which is acceptable-it can accept the language of two or more parties to transfer to the other language and the same It is the community semantic management system of Tong-Chu Lei Acting Authority; this invention conforms to the requirements of the knowledge economy, so it is a basic tool of the knowledge economy.
運用上述方法可用語態詞找出相關之回答文章中可能用的語意 語態詞對應;如:有多少/種類/ V樹木?有多少對應有/還有/ 及/等詞目,種類/樹木為名詞(詞性符射看出)也是找資料所需反 過來用的現用之名詞’此用問的詞類來搜尋回答所需的句構中各種詞類 的來源’加上相對應的語態詞,作成人事時地物所用的語態詞,作一排 列表’成為各企業文檔所服務的内容表,作為企業問答題互動題庫,即 當有問題者其間之觸含語態贿題庫内容產生交騎應_後;可以 不用-般型式之專業版之逐—反問各種條件情況給問者喃,而可以將 所缺之相關之主題及語態文法,反問即可,此用於知識搜尋與專業問題 回答,有快速精準的智慧處理功用。Using the above method, you can find out the possible semantic meanings in the relevant answer articles by using the voice words; for example: how many / type / V trees? How many words are there / and / and /, and the category / tree is a noun (seeing part-of-speech), which is also the current noun used in reverse to find the information. The source of various part-of-speech in the sentence structure, plus the corresponding modal words, as the modal words used by adults as current features, make a row of lists, become the content table served by each corporate document, and serve as an interactive question bank for corporate questions and answers. That is, when the content of the questionable bribery question bank during the question is generated, you can use the professional version of the general type instead of asking the questioner to whisper all kinds of conditions, and you can use the relevant topics that are missing. And voice grammar, just ask back, this is used for knowledge search and professional question answering, with fast and accurate intelligent processing function.
以上方法對於本發明之原理敘述或為闡明解決方案之目的,其各式 符號或流程位置順序僅是舉例運算用途之一,而無意限定本發明精確地 揭露原理方法,故依各思考平台流程與審查同模組文法碼對應資料,可 讓熟習該項技術者以各種程序符號定義但用思考平台模式而達到實際 上語意運异效果,本發明的技術思想由以下的申請專利範圍及其内容項 目來決定。 1237188 讀 【圖式簡單說明】 對於本技藝之人式而言,從以下所做的表格建構步驟及流程程序, 本發明將能夠清楚的被了解,上述原理方法及目旳優點將會更加明 示,其中: 圖一、文字詞構平臺 例句APO圖示說明補充The above method describes the principle of the present invention or clarifies the purpose of the solution. The various symbols or the sequence of the process positions are only one of the purposes of example calculations. It is not intended to limit the present invention to accurately disclose the principle and method. Examining the corresponding data of the grammar code of the same module can allow those skilled in the technology to use various program symbols to define but use the thinking platform mode to achieve the actual semantic difference effect. The technical idea of the present invention is covered by the following patent application scope and its content items To decide. 1237188 Read [Schematic description of the figure] For the person of this skill, from the following table construction steps and process procedures, the present invention will be clearly understood, the above principles and methods and objectives advantages will be more clearly shown, Among them: Figure 1. Text word structure platform example sentence APO icon description supplement
圖二、基因碼組成及文法符號定義 圖二、字排框表 圖四、基元決策圖(含例外表) 圖五、文字基本邏輯平面 圖六、片語文法碼組合對應 圖七、跨片語組合邏輯 圖八、片語子句的組合邏輯 圖九、語態子句組合邏輯模組 圖十、句型定義邏輯模組 圖十一、句構文法符號組合流程圖 圖十二、單APO近似模組表 圖十二之一 圖十三、AQ表 19 1237188 圖十四、答案計算表 【主要部分代表符號說明】 1.問句及答案句詞構基因碼 2.子句詞構基因碼模組 3.邊除化基因碼模組 4.主要句子基因碼模組 5.組合基因碼模組Figure 2. Gene code composition and grammatical symbol definition. Figure 2. Font box table. Figure 4. Primitive decision chart (including exception table). Figure 5. Basic logical plan of the text. Combining logic diagram 8. Combining logic diagram of phrase clauses 9. Combining logic module diagrams of modal clauses. 10. Logic module diagram of sentence pattern definition. 11. Combining flowcharts of grammatical symbols of sentence structure. 12. Single APO approximation. Module table chart one twelfth figure AQ table 19 1237188 Figure fourteen, answer calculation table [Description of the main parts of representative symbols] 1. Question and answer sentence word formation gene code 2. Clause word formation gene code pattern Group 3. Edge Deletion Gene Code Module 4. Main Sentence Gene Code Module 5. Combination Gene Code Module
2020