1226560 捌、本案若有化學式時，請揭示最能顯示發明特徵的化學式 玖、發明說明： 【發明所屬之技術領域】 本發明係關於一種具自然語言解析能力的資訊系統及 處理方法’尤指一種利用自然語言解析手段以直接解構資 讯耑求者以自然語言形式表達的需求内容，進而與以相同 技術構成的資料庫内容進行媒合，而以媒合度最高的内容 回應予 > 料需求者，藉此，令搜尋資訊之人機介面更具親 和力。 【先前技術】 現代人對於網際網路的形容詞通常是一座浩翰無際的 二貝料庫，其内容之豐富直令人取之不盡用之不竭，而為方 便使用者搜尋資料，各大入口網站均有提供搜尋者 ^ 田 使用者在搜尋攔位輸入關鍵字並執行搜尋，即可找出眾多 與該關鍵字有關的網站、網頁或文章等。但這是不是一種 有效貫用的方法是值得討論的，事實上，對於一個熟習電 腦操作及深諳網路資料庫結構的人來說，搜尋引擎不失為 一種實用的方式，因為其僅得如何下關鍵字，即使搜尋出 來的資料量十分龐大，亦具備一定的技巧去抽絲剝蘭地找 出舄要的 > 料然而，對於不具備前述知識技能的使用者 而言，想要透過搜尋引擎快速的找到所需的資訊，是需求 1226560 碰運氣的。 事實上，現代科技愈來愈講究人機介面的親和力，換 言之，是儘其可能的降低操作技巧的成分，卻能相對提高 操作的準確性與有效性，因而即有所謂的人工智慧系統= 問世，這些系統之目的無非是為了使用者更方便操作。以 前述資訊搜尋方法而言，最直接的方法莫過於使用者直接 以自然語言表達其需求，而系統本身可以經由對自然語言 =解析，瞭解其需求與搜尋内容，進而找出符合其需求的 資料予以回應，例如使用者說了或輸入“這麼胖，怎麼辦 ?，此時系統就會自動判斷使用者是希望找到減肥的方 法或相關資料，在解析出使用者的期望後，即可直接找出 與減肥有關的資料予以回應。然而，#想達到前述目的， 必須有相當先進的技術支援，否則前述狀況亦只是理想而 已〇 而既有專利文獻所揭露關於，，自然語言，，的相關技術， 大致可分為以下兩種·· 一種是統計及機率的方式，另一種則是〇nt〇丨方式 、。但採統計及機率的方式，就目前已知的技術，其準確率 並不间，原因在於其沒有一解析機制，亦沒有配合運作的 知識庫，而造成其準確率偏低。 再者，ontology方式的缺點在於需要投入大量人力以 建置知識庫，即使如此，因。ntology未針對句子的文法進 行剖析，所以對於長句與複合句均無法有效地理解。 故由上述可知，自然語言應用於網路資源的搜尋，雖 !226560 可提升人機介面的親和力，但就現有技術而言，準確率偏 低、缺乏有效的解析機制及建制知識庫須耗費大量資金等 問題，均有待進一步克服解決。 【發明内容】 因此，本發明主要目的在提供一種具自然語言解析能 力的資訊處理方法，其可供使用者直接以自然語言表達其 搜尋資料之意思，經由一自然語言解析手段直接解構其需 求内容後，即與資料庫内容進行媒合，而以媒合度最高的 2容回應予資料需求者；藉此可透過強大的語言解析能力 提南資訊搜尋的準確率，而方便使用者取得需要的資料， 另以相同技術建立知識庫則可大幅降低建置成本。 為達成前述目的採取的主要技術手段係令前述方法包 括下列步驟： 輸入自然語言之詞句； 執行-自然語言解析手段，以便將自然語言轉換成一 具有特&事件背景、需求條件的“建構式概念腳本，格式 執行一媒合手段， 一資料庫以相同技術產 係令所產生“建構式概念腳本，，與 生的“建構式概念腳本，，進行媒合 取媒合度最高資料回應需求 在前述方法中， 產生一反應特定事件 、過對自然語言的逐步解析，可據以 特疋條件的“建構式概念腳本，，， 1226560 二:和資料庫内容進行媒合，巾資料庫内容亦係以相同 的技術將眾多知識轉換成“建構式概念腳本，，，並同時記 錄其回應資料，故在媒合步驟中，即由新產生@ “建構式 概念腳本”與資料庫中的“建構式概切本，，冑行媒合， 利用邏輯判斷以找出相同或相近的資料内纟，藉此可供使 用者以自心言描述其需求’而提供更方便的資訊搜尋方 法0 刖述自然語言解析手段包括下列步驟： 檢查句型，確認輸入詞句屬於提出需求之語言句型； 斷詞，係對輸入詞句進行斷詞； 專業領域分類，係用以賦予斷詞後每一字詞之專業屬 I*生’如區分為專業詞、一般詞或新詞等； ' 關鍵詞組檢查，由需求問句中檢查是否存在顯示其需 求核心之關鍵詞組； 同義詞或同義詞組檢查，檢查需求問句中是否存在專 業巧之同義詞或關鍵詞組之同義詞組； 轰生一代表使用者需求的“建構式概念腳本” (Constructive Concept Script)。 前述自然語言可以語音方式說出，經語音辨識技術處 理後’再進行自然語言解析。 前述的句型檢查係透過一句型比對技術所達成。 前述專業領域分類係與一詞庫之内容進行比較，如為 0司庫中具有的專業詞，即定義為專業詞，如不是，則判斷 其是否詞庫中的一般詞，如是即定義為一般詞，不是則定 1226560 義為新詞。 -定立’係針對每-字詞在特定領域與 現頻率，以計S二現頻率及在某-篇文章中的出 其區分為”專掌二的權重’再根據權重分數高低將 义 …习（D〇mam)及，，一般詞"（generic)。 則述的媒合手段包括下列步驟： 建構式概念腳本”； 與資料庫中搜尋到的 在資料庫中搜尋相同或近似的 “令需求者的“建構式概念腳本 “建構式概念腳本，，進行邏輯判斷 依媒合度高低提供解析式回應。 建構式概“本，，内容包括―“關鍵事件，，輿 一 條件’；其中： 該“關鍵事件”下具有複數的關鍵詞組； 該“條件，，下亦具有複數的關鍵詞組，各關鍵詞組以 下仍分別具有複數的詞組。 前述的‘‘建構式概念腳本，，搜尋係由下列步驟組成： 以需求“建構式概念腳本”中的專業詞去搜尋資料庫 中各個“建構式概念腳本，，的專業詞詞庫； 根據搜尋到的專業詞，在資料庫中找出相關的關鍵詞 組（N-Gram); 根據找出的關鍵詞組進-步搜尋資料庫中所有相關的， 關鍵事件”、”條件”； “根據搜尋到所有相關的”關鍵事件，，、，，條件，，找出可能的 建構式概念腳本”。 1226560 ^前述“建構式概念腳本，，搜尋步驟中，於找出專業詞 後1針對專業詞進—步找出其同義詞；又在找出關鍵詞 組後，亦進一步找出關鍵詞組的同義詞。 前述的邏輯判斷係根據兩個“建構式概念腳本，，内容 的差異以進行比對。 刖述邏輯判斷係根據兩個“建構式概念腳本，，内的子 集合之間不同的差異，定義了 ”EQU”、"MAX”、，，M|N”及 ”x〇R”等四個運算子，以計算出各個不同‘‘關鍵詞組，，子 集：間之關係，並可進一步推導出各個不同“關鍵事件” 、“條件”子集合間之關係’最後據以找出可能的“建構 式概念腳本”。 本發明又-目的在提供一種具自然語言解析能力的資 訊系統，其包括有： 一“建構式概念腳本，，資料庫，係透過一自然語言解 析手段將眾多知識轉換成“建構式概念腳本，]各式並予儲 存； -使用者介面，供使用者以自然語言輸人其對於相關 資訊之需求，纟同樣以自然語言解析手段將輸入的自然語 言轉換成“建構式概念腳本，，袼式； “媒口機制’係、針對使用者“建構式概念腳本，，與資 料庫“建構式概念腳本，，進行媒合； -邏輯解譯單it，係針對前述媒合機制產生的結果進 行分析’隨即可回應予使用者。 前述系統進-步包括一後端管理機制，用以處理評分 8 1226560 不局的“建構式概念腳本，，，藉以補強擴充資料庫内容。 主要係針對評分不高或使用者不滿意回覆的“建構式概念 腳本交由前述後端管理機制處理，該後端管理機制至少 包括： 一確認管理介面，係供確認依使用者需求所提“建構 式概念腳本，，之問題是否已存在資料庫中，且分數是否為 可接受範圍内，若是，則不需要再作其他處理，若不是， 則送至下一管理介面； 一同義詞或同義詞組檢查/產生介面，係根據觀察問 句中疋否有同義詞或同義詞組干擾了問句媒合的準確度， 透過同義詞/同義詞組的建立，並和既有詞庫建立關連， 以排除同義詞或同義詞組的干擾’進而提升媒合的準確度 字檢查等：：==對使用者之問句進行精簡、錯 動作！過正規化之後，即重新進行媒合，如 口又：坪分為可接受範圍，即無須再進行其他處理；、 新詞:===斷詞後開始檢查問句中是否有 ，並賦予詞性:在資料庫中增加—筆新詞 W二ϋ校f管理介面，係用以檢查專業詞與-般詞之 j仙讀正問句巾料㈣及-被誤判 如確遇為誤判，即予重新定義，並對問句 既人 本，，重新進行媒合。 構式枝必腳 1226560 【實施方式】 本發明之自然語言解析技術，主要可供使用者直接以 自然語言表達其對於尋找相關資訊的意願與需求内容，再 經由系統的自然語言解析過程，即可將其轉換成一種特殊 的“建構式概念腳本”格式，再與資料庫中以相同技術轉 換而成的眾彡建構式概念腳本，，進行媒合，以找出最符 合使用者需求的相關資訊，例如有個人只說了 ： “天氣這 麼熱’ a又胖到1 0 〇公斤，不知道怎樣才能瘦一點？” ，由前述語意中反映出其尋找減重方法的需求，此時本發 明的系統與方法可以自動解析判斷其語意，進而瞭解其需 求，再依其需求找出相關的資訊，更特別的是，在解析其 語意的過程中，可以自動排除與其真義無關的贅詞、贅字 ，以前述語言為例，重點係在於重達1 0 0公斤的人如何 找到適合的減重方法？至於天氣冷熱應該不是重點，故可 予忽略此種方式，可供使用者以自然語言表達其想要 找到的> Λ @本發明亦得直接解析其語意，以滿足其需 长至於達成刖述目的之具體技術内容，詳如以下所述： 如帛圓所不，係本發明之系統架構示意圖，其包括 有： =自‘然語言解析能力的資訊系統，其包括有： 、 建構式概念腳本”資料庫（1 〇 )，係透過-自 然語言解析手段（1 1 ) ^ j將眾多知識轉換成一種特殊的“ 建構式概念腳本，，格式並予儲存； 使用者’丨面（2 〇 )，供使用者以自然語言輸入其 1226560 對於相關資訊之需求，並同樣以一自然語言解析手段（2 1 )將輸入的自然語言轉換成“建構式概念腳本”格x式； -媒合機制（3 0) ’係針對使用者“建構式概念腳 本與資料庫建構式概念腳本”進行媒合· -邏輯解譯單元（4Q)，係針對前述媒合機制產生 的結果進行分析，其結果將送至電腦系統的調度程式 (Dispatcher) (4 1)以回應予使用者；其中： 前述的自然語言解析手段（工工）係以軟體達成，其 工作流程係如第二圖所示，包括下列步驟： 檢查句型（201) ’確認輸入詞句屬於提出需求之語言句 型，其一般為問句或祈使句，此步驟係用以確認使用者提 出需求之意願，當其輸入自然語言詞句符合特定句型時， 即認定其確有搜尋資訊之需求，故執行下一步驟； 斷词（202) ’係對輸入詞句進行斷詞； 專業領域分類（203)，係用以賦予斷詞後每一字詞之專 業屬性/權重，如區分為專業詞、一般詞或新詞等；其進 一步的詳細技術手段容後詳述； 、關鍵詞組檢查（2〇4)，由問句中檢查是否存在顯示其需 2核心之關鍵詞組，其大致分為兩類：一種是代表某種特 疋事件或背景，另一種代表該資訊的各種相關“條件，，； 同義詞或同義詞組檢查（2〇5)，係檢查問句中是否存在 專業阔的同義詞或關鍵詞組的同義詞組，其中同義詞檢杳 係找出專業詞的同義詞，同義詞組檢查則找出與關鍵詞 的同義詞組； / 1226560 產生一代表使用者需求的“建構式概念腳本，， (Constructive Concept Script)(206)。其中： 則述建構式概念腳本”的内容請參閱第三圖所示， 其包括兩大群組，一為關鍵事件（Key event)、另一為“條 件” （condition);其中，關鍵事件以丁包括一句型格式 (Sentence)及多數與事件有關的關鍵詞組（N Gram)，每一 關鍵詞組（N_G「am)之下又包括多數的触㈣阳e)，而構 成Μ狀π構。X冑件，，群組下仍具備由多數關鍵詞組 (N-Gram)、詞組（Phrase)組成的樹狀結構内容。 刖述的句型檢查係透過一句型比對技術所達成。 前述專業領域分類係與一詞庫之内容進行比較，如為 詞”具有的專業詞1定義為專業詞，如不是，則判斷 其疋否詞庫中的一般詞，如是即定義為一般詞，不是則定 義為新詞。 一=於4專業詞庫的建立，係針對每—字詞在特定領域 、疋數量以上的文章中的出現頻率及在某一篇文章中的 =頻率’以計算出該字詞的權重，再根據權重分數高低 將其區分^專業詞，，(Domain)及”一般詞，，(gene「ic)。 • 述的媒ϋ手段係如第四圖所示，包括下列程序 _)在資料庫中搜尋㈣或近似的“建構式概念腳本” #者&建構式概念腳本”肖資料庫中搜尋到的 冓式概念腳本，，進行邏輯判斷（402); 12 1226560 依媒合度高低提供解析式回應（403)。其中： 前述的“建構式概念腳本，，搜尋方式係如第五圖所示 步驟達成··（其工作流程請配合參閱第六圖所示） 以新產生建構式概念腳本”中的專業詞去搜尋資料 庫中各個建構式概念腳本”的專業詞詞庫（5〇 1) ,· 根據搜尋到的專業詞，在資料庫中找出相關的關鍵詞 組（N-Gram)(502);1226560 若 If there is a chemical formula in this case, please disclose the chemical formula that can best show the characteristics of the invention 玖 Description of the invention: [Technical field to which the invention belongs] The present invention relates to an information system and processing method with natural language analysis capabilities, especially a Use natural language analysis to directly deconstruct the information content demanders express in natural language, and then mediate with the content of the database composed of the same technology, and respond with the content with the highest match to the material demander. To make the human-machine interface for searching information more accessible. [Previous technology] Modern people ’s adjectives for the Internet are usually a vast and expansive database of two shellfish. Its rich content is inexhaustible. For the convenience of users to search for data, major The portals all provide searchers ^ users who enter keywords in the search block and perform searches to find many websites, web pages or articles related to the keywords. But it is worth discussing whether it is an effective and consistent method. In fact, for a person who is familiar with computer operation and deep knowledge of the network database structure, the search engine is a practical way because it only has to be done. Keywords, even if the amount of data searched is very large, they still have certain skills to find the most important information. However, for users who do not have the aforementioned knowledge and skills, they want to use search engines Finding the required information quickly requires 1226560 for luck. In fact, modern technology pays more and more attention to the affinity of human-machine interface, in other words, it reduces the components of operation skills as much as possible, but can relatively improve the accuracy and effectiveness of operations, so there is a so-called artificial intelligence system = come out The purpose of these systems is nothing more than user convenience. In terms of the aforementioned information search methods, the most direct method is that users directly express their needs in natural language, and the system itself can understand its needs and search content through analysis of natural language =, and then find data that meets its needs Respond, for example, the user said or typed "So fat, what should I do? At this time, the system will automatically determine whether the user wants to find a way to lose weight or related information. After analyzing the user's expectations, he can directly find To respond to the information related to weight loss. However, #To achieve the aforementioned purpose, there must be quite advanced technical support, otherwise the aforementioned situation is only ideal. Existing patent documents disclose related technologies related to natural language. It can be roughly divided into the following two types: one is the method of statistics and probability, the other is the method of ntnt. However, the method of statistics and probability is based on currently known technologies, and its accuracy is not constant. The reason is that it does not have an analysis mechanism or a cooperative knowledge base, which results in a low accuracy rate. Furthermore, ont The disadvantage of the ology method is that it requires a lot of manpower to build a knowledge base. Even so, because ntology does not analyze the grammar of the sentence, it cannot effectively understand long sentences and compound sentences. From the above, we can see that natural language applications Searching for network resources, although! 226560 can improve the affinity of the human-machine interface, as far as the existing technology is concerned, the problems of low accuracy, lack of effective analysis mechanism, and the need to build a large amount of knowledge base need to be overcome. [Summary of the invention] Therefore, the main object of the present invention is to provide an information processing method with natural language parsing capability, which can allow users to express the meaning of their search data directly in natural language and directly deconstruct it through a natural language parsing method. After the content is requested, it will be matched with the content of the database, and respond to the data requester with the highest matching capacity; this can improve the accuracy of the information search through powerful language analysis capabilities, and facilitate users to obtain the needs Data, and building a knowledge base with the same technology can greatly reduce the The main technical means adopted to achieve the aforementioned purpose is to make the aforementioned method include the following steps: input the words and expressions of natural language; execute-natural language parsing means to transform natural language into a "constructive form with special & event background and demand conditions" Concept script, format executes a matchmaking method, a database uses the same technology production line to generate a "constructive concept script," and a "constructive concept script" to perform matchmaking to obtain the highest matchmaking data. In the method, a specific event is generated, and a gradual analysis of natural language is generated. The "constructive concept script" based on special conditions can be used to match the contents of the database. The contents of the database are also based on The same technology converts a lot of knowledge into "constructive concept scripts, and records their response data at the same time, so in the matching step, a new @constructive concept script" and the "constructive concept in the database" Cut and match, use logical judgment to find the same or similar data content, so it can be used Provide a more convenient information search method by describing your needs with your own words. 0 Narrative natural language analysis methods include the following steps: Check the sentence pattern to confirm that the input words and sentences belong to the language sentence type that raises the requirements; Word segmentation is performed on the input words and sentences. Word segmentation; professional field classification, which is used to give each word after the word segmentation to the professional genus I * sheng, such as distinguished into professional words, general words or new words, etc .; There are keyword groups that show the core of their needs; synonym or synonym group checks to check whether there are professional or synonym groups in the demand question; a "Constructive Concept Script" (Constructive Concept) that represents user needs Script). The aforementioned natural language can be spoken in a voice manner, and then processed by the speech recognition technology 'before performing natural language analysis. The aforementioned sentence pattern check is achieved through a sentence pattern comparison technique. The foregoing professional field classification system is compared with the content of a thesaurus. If it is a professional word in the 0 treasury, it is defined as a professional word. If it is not, it is judged whether it is a general word in the thesaurus. If it is, it is defined as a general word. If not, then 1226560 is defined as a new word. "Dingli" is based on the specific frequency and present frequency of each word in the specific field, taking the second frequency of S and the distinguishing in a certain article into "weights in charge of two", and then according to the weight scores ... (D〇mam) and, the general term " (generic). The matching means described below include the following steps: Constructive Concept Script "; Searching in the database for the same or similar" command " The demander's "constructive concept script" constructs a conceptual concept script and makes logical judgments to provide an analytic response based on the degree of media integration. The "constructive profile", which includes "" critical events, a condition "; of which: There are plural keyword groups under "key event"; under the "condition, there are also plural keyword groups below, and there are plural phrases below each keyword group. The aforementioned "constructive concept script" search consists of the following steps: search for each "constructive concept script" in the database using the professional words in the "constructive concept script"; according to the search The relevant professional words are found in the database and related keyword groups (N-Gram) are searched. Based on the found keyword groups, all relevant and key events in the database are further searched. All relevant "key events ,,,,, conditions, find possible constructive concept scripts". 1226560 ^ In the aforementioned "constructive concept script," in the search step, after finding the professional words, 1 for the professional words— Step by step, find out its synonyms; after finding out the keyword group, further find out the synonyms of the keyword group. The aforementioned logical judgment is based on the differences between the two "constructive concept scripts, and the contents are compared. The logical judgment is based on the differences between the two sub-sets within the two" constructive concept scripts, " "EQU", " MAX ", ,, M | N", and "x〇R" and other four operators to calculate the different `` keyword groups, '' subsets: the relationship between them, and can be further derived The relationship between the different "key events" and "conditions" subsets was finally used to find possible "constructive concept scripts." Another object of the present invention is to provide an information system with natural language parsing capabilities, including: a "constructive concept script, a database, which transforms a large amount of knowledge into a" constructive concept script through a natural language parsing method, ] Various types and storage;-User interface for users to input their needs for relevant information in natural language, and also use natural language analysis to convert the input natural language into "constructive concept script, ; "Matching mechanism" is a "constructive concept script for users", and a "constructive concept script for a database" is used to match;-a logical interpretation sheet it, which analyzes the results generated by the aforementioned matchmaking mechanism 'You can then respond to the user. The aforementioned system further includes a back-end management mechanism to handle the "constructive concept script 8" which does not end with the score 8 1226560. It is used to reinforce and expand the content of the database. It is mainly aimed at "scores that are not high or users are not satisfied with the response" The constructive concept script is handed over to the aforementioned back-end management mechanism. The back-end management mechanism includes at least: a confirmation management interface for confirming whether the "constructive concept script" according to the user's needs has been stored in the database. And whether the score is within the acceptable range, if it is, then no further processing is needed, if not, it is sent to the next management interface; a synonym or synonym check / generation interface is based on whether there is any in the observation question Synonyms or synonym groups interfere with the accuracy of question matching, through the establishment of synonyms / synonym groups, and the connection with the existing thesaurus, in order to eliminate the interference of synonyms or synonym groups, and thus improve the accuracy of matching ：: == Simplify the user ’s question and make mistakes! After normalization, rematching will be performed again. Acceptable range, that is, no further processing is needed; New words: === Begin to check if there are questions in the sentence after segmentation, and give part-of-speech: add in the database-a new word W 二 ϋ 校 f management interface, It is used to check the professional words and the general words of the j-sentence of the positive question sentence, and-if it is misjudged, it is redefined if it is misjudged, and the question is both human and re-matched.式 枝 必 脚 1226560 [Embodiment] The natural language parsing technology of the present invention mainly allows users to directly express their wishes and requirements for finding relevant information in natural language, and then through the system's natural language parsing process, It is transformed into a special "constructive concept script" format, which is then matched with popular constructive concept scripts converted from the database using the same technology, and matched to find the relevant information that best meets the needs of users. For example, one person only said, "The weather is so hot, 'a is fattening to 100 kilograms. I don't know how to get thinner? ”, The need to find a weight-loss method is reflected in the foregoing semantic meaning. At this time, the system and method of the present invention can automatically analyze and judge its semantic meaning, and then understand its needs, and then find relevant information according to its needs, more particularly, In the process of parsing its meaning, it can automatically exclude redundant words and words that are not related to its true meaning. Taking the aforementioned language as an example, the focus is on how to find a suitable weight loss method for people weighing 100 kilograms. As for the hot and cold weather It should not be the point, so this method can be ignored, and the user can express what he wants to find in natural language. Λ @The present invention must also directly analyze its semantic meaning to meet the specific needs that it takes to achieve the stated purpose. The technical content is as follows: As shown in the following figure, it is a schematic diagram of the system architecture of the present invention, which includes: = Information system with natural language analysis capabilities, which includes: (1 〇), through a natural language parsing method (1 1) ^ j to transform a lot of knowledge into a special "constructive concept script, formatted and stored; make User's face (20), for users to input their 1226560 needs for related information in natural language, and also use a natural language analysis method (2 1) to convert the input natural language into "constructive concept script" Lattice x-style; -Matching mechanism (30) 'Matches the user's "constructive concept script and database constructive concept script" · -Logical interpretation unit (4Q), which is generated for the aforementioned matchmaking mechanism The results are analyzed and the results will be sent to the Dispatcher (4 1) of the computer system to respond to the user; of which: the aforementioned natural language analysis method (engineering) is achieved by software, and its workflow is as follows As shown in the second figure, the following steps are included: Check the sentence pattern (201) 'Confirm that the input phrase is a language sentence pattern that raises a demand, which is generally a question or an imperative sentence. This step is used to confirm the user's willingness to make a request. When the input natural language phrase matches a specific sentence pattern, it is determined that it does have a need to search for information, so the next step is performed; word segmentation (202) 'is to segment the input phrase; specialty Domain classification (203) is used to assign professional attributes / weights to each word after the word segmentation, such as distinguishing into professional words, general words or new words; its further detailed technical means will be described later in detail; Check (204) checks whether there is a keyword group showing that it needs 2 cores in the question sentence, which can be roughly divided into two categories: one represents a specific event or background, and the other represents various related information of the " Condition, ;; Synonym or synonym check (205) is to check whether there is a specialized broad synonym or keyword group in the question. The synonym check is to find the synonym of the professional word, and the synonym check is to find Generate a synonym group with keywords; / 1226560 Generate a "Constructive Concept Script" (206) that represents the needs of the user. Among them: Please refer to the third figure for the content of the constructive concept script. It includes two groups, one is a key event and the other is a condition. Among them, the key event is Ting includes a sentence pattern (Sentence) and most event-related keyword groups (N Gram), and each keyword group (N_G "am) includes a majority of touching the positive e), forming an M-shaped π structure. X For example, the group still has a tree-like structure composed of most keyword groups (N-Gram) and phrases (Phrase). The described sentence pattern check is achieved through a sentence pattern comparison technique. The aforementioned professional field classification It is compared with the content of a thesaurus. If the professional word 1 in the word "is defined as a professional word, if it is not, it is judged whether it is a general word in the dictionary. If it is, it is defined as a general word. If not, it is defined as new word. 1 = The establishment of the 4 professional thesaurus, which is based on the frequency of occurrence of each word in a specific field, more than the number of articles, and = frequency in an article to calculate the weight of the word, and then Differentiate them according to their weighting scores ^ professional words, (Domain) and "general words," (gene "ic). • The means of matchmaking are shown in the fourth figure, including the following procedures _) in the database Search for “constructive concept script” or similar “constructive concept script” # 者 & constructive concept script ”to search for the conceptual concept script in the Xiao database, and make logical judgments (402); 12 1226560 provide analytic response according to the level of media (403). Among them: The aforementioned "constructive concept script, the search method is achieved as shown in the fifth figure ... (for its workflow, please refer to Figure 6 for cooperation) Use the professional words in the newly generated constructive concept script" Search the professional word thesaurus for each constructive concept script in the database (501), · According to the searched professional words, find the relevant keyword group (N-Gram) in the database (502);
根據找w關鍵詞組進一步搜#資料冑中所有相關的” 關鍵事件”、”條件”(503); 根據搜尋到所有相關的，，關鍵事件”、，，條件，，找出可能的 建構式概念腳本”（504)。 前述“建構式概念腳本，， 後’將針對專業詞進一步找出 組後’亦進一步找出關鍵詞組 準確性。 搜尋步驟中，於找出專業詞 其同義詞；又在找出關鍵詞 的同義詞組，以提高比對的 在完成前述搜尋步驟後’即進一步進行邏輯判斷，According to the search keyword group, further search all relevant "key events" and "conditions" (503) in the data set; according to all relevant, key events, and conditions, find possible constructive concepts Script "(504). In the aforementioned "constructive concept script, the term" will further find the group after the professional word "also further finds the accuracy of the keyword group. In the search step, the synonyms of the professional word are found; the synonyms of the keyword are also found In order to improve the comparison, after completing the foregoing search steps, 'there is a further logical judgment,
=七圖所示，該邏輯判斷係根據—“建構式概念腳本 内的子集合之間不同的差異，定義了”EQu” 吉子集合間之關係’並可進一步推導出各個不同 “條件”子集合間之關係’最後據以找出可能 六建構式概切本，，。由s七圖可看出邏輯判斷之工作 谷，其揭示新產生的“建構式概念腳本” （γ 〇 )欠 庫中眾多“建構式㈣腳本，，（8〇)進行邏輯判斷^ 13 1226560 二由明顯看出，資料庫中的每-“建構式概念 腳本（8 0 )係分別與複數的‘‘關鍵事件，，（8 i )、 事件（8 2 )關連，每一‘‘關鍵事件” （8工）、“ 事件（8 2 )又分別與複數的“關鍵詞組，，（8 3 )( 8 4)關連’每-“關鍵詞組，，（ 8 3 ) ( 8 4 )再盥複 數的詞組（8 5 ) ( 8 6 )關連。 八 而如第八圖中虛線框所代表的物件（8 〇 i )〜（8 〇 η) ”’其集合在第九圖中可以得到由新產生“建構式概 念腳本’’ （7 〇 )到每一個物件（8 〇丄）〜（8 〇 η ) 之間的關，透過此從屬關係進行評分，即可得到該“建 構式概念腳本”的媒合度分數。 而系統亦將根據前述的媒合結果依評分狀況，回應予 使用者’由下列的第十圖吾人可以歸納出一些原則，透過 比較新產生“建構式概念腳本，，與媒合度最高“建構式概 念腳本”《間的關係，可以比數的方式描述其間的相對關 係，如：. XX:yy，其中xx代表新產生的“建構式概念腳本” ’ yy則代表媒合度評分最高的“建構式概念腳本”，其可 能分析結果可如下列： 1 〇〇 : 1 〇〇 :您的問題内容系統完全理解，系統也有 學過S以下是為您的搜尋到適合的答案。 70以上：1 〇〇 ··您的問題内容系統有些不是很清楚， 系統年紀還小，所學過的知識可能無法完全滿足您的需求 ，以下是系統能找到最接近的解答。 1226560 100 : 70以上 系統學過的知識可能^ 找到更詳盡的解答。 •态的問題内容系統完全理解，但是 比您的問題内容更深入，以下是系統 10 0 : 7 0以下•你^ •心、的問題内容系統完全理解，但是 系統學過的知識可能4彳尔Μ 』此比J的問題内容更深入且涵蓋更廣， 以下疋系統找到更詳盡的解答。 、下· 1 00 ·您的問題内容系統报多都不 ，系統具備的知識可能盔法〜入？+ J月b…在元全滿足您的需求，以下是系= As shown in figure 7, the logical judgment is based on "the difference between the sub-sets in the constructive concept script, defining the relationship between the" EQu "and the sub-sets," and further deducing the different "condition" sub-sets. Finally, we can find the possible six constructive sketches. The work valley of logical judgment can be seen from the seven diagrams, which reveals the newly generated "constructive concept script" (γ 〇) in the library. Many "constructive㈣scripts, (8〇) make logical judgments ^ 13 1226560 It is obvious that every-" constructive concept script (8 0) in the database is separate from the plural "key events, (8 i), event (8 2), each "key event" (8 workers), "event (8 2) is respectively related to the plural" keyword group, (8 3) (8 4) ' Each- "keyword group, (8 3) (8 4) is related to the plural phrase (8 5) (8 6). Eight and the objects represented by the dashed box in the eighth figure (80) ~ ( 8 〇η) "The collection can be obtained in the ninth figure by the newly generated" constructive concept script " (7 〇) to each object (80 〇) ~ (80 〇), through this affiliation to score, you can get the "constructive concept script" match degree score. And the system will According to the aforementioned matchmaking results and responding to the users according to the scoring status, we can summarize some principles from the tenth figure below. By comparing the new generation of "constructive concept scripts," the "constructive concept script" with the highest degree of matchmaking " The relationship between them can be described in a comparative way, such as: XX: yy, where xx represents the newly generated "constructive concept script", yy represents the "constructive concept script" with the highest match score, The possible analysis results can be as follows: 〇〇: 1 〇〇: Your question content system is fully understood, the system has also learned S The following is a suitable answer for your search. 70 or more: 1 〇〇 ·· The content of your question system is not clear, the system is still young, and the knowledge you have learned may not fully meet your needs. The following is the closest solution that the system can find. 1226560 100: 70 or more system knowledge may find more detailed answers. • The content of the problem system is fully understood, but it is more in-depth than your problem content. The following is the system 10 0: 7 0 or less. • You ^ • The content of the problem and the system is fully understood, but the knowledge learned by the system may be 4 彳. This is more in-depth and broader than J's question. The following 内容 system finds a more detailed answer. · Next · 1 00 · Your question content system is not reported too much, the system possesses knowledge that may be helmeted ~ enter? + J 月 b… At Yuanquan to meet your needs, the following is the department
統能找到最接近的解答。 99〜70: 99〜70:您的問題内容系統大致理解，但是 系統具備的知識可能和您的問題有些許的出入，以下是系 統找到最接近的解答。 69 40 · 69〜40 :您的問題内容系統大部分尚未接觸 過，而且系統具備的知識可能和您的問題有蠻大的差距， 以下是系統找到最可能的解答。The system can find the closest solution. 99 ~ 70: 99 ~ 70: The content of your question is generally understood by the system, but the knowledge possessed by the system may differ slightly from your question. The following is the closest solution the system has found. 69 40 · 69 ~ 40: Most of the content of your question system has not been contacted, and the knowledge possessed by the system may be quite different from your question. The following is the most likely solution that the system can find.
40以下 40以下：您的問題内容系統可能沒接觸過， 系統將盡快擴充新知，透過您的問題和互動會讓系統更加 聰明，以下是系統找到最相關的解答。 〇 : 0 : 100 :很抱歉，您的問題内容系統可能沒接觸 過’或是您的問題内容未清楚說明。以下是系統搜尋出可 能有相關的答案。 由上述說明可知’本發明利用自然語言解析手段為核 心所提供的資訊系統’其可方便使用者利用自然語言提出 其對於資訊的需求，所謂的自然語言，其可以是文字，亦 15 1226560 可以是語音，當語音方式說出，只須經語音辨識技術處理 後，再進行自然語言解析即可達成前述實施例相同之效果 Ο 由於資訊系統應不斷吸收新知，以滿足更多使用者的 需求’其意味著使用者提出的資訊需求是系統無法作答， 或作答評分不冑’另如使用者對回應狀況不滿意者，為解 決是項_，本發明令前㈣統進—步具有—後端管理機 制’其第-階段係提供下列管理介面，以便將媒合評分偏 低及使用者不滿意的回應送至此―機制進行處理，如第十 一圖所示，其包括有： 確涊官理介面（5 1 )，係供確認依使用者需求所 提胃“建構式概切本” t問題是否已存在資料庫中，且分 數是否為可接受範圍内，若是，則不需要再作其他處理， 若不是，則送至下一管理介面； /正規化管理介面（52),係對使用者之問句進行 精簡、錯字檢查等動作，經過正規化之後，即重新進行媒 合，如媒合度之評分為可接受範圍，即無須再進行其他處 理； 一新詞檢查介面（5 3 ),係在斷詞後開始檢查問句 中是否有新詞出現，如經認定為新詞，即在資料庫中增加 一筆新詞，並賦予詞性； 一定義校正管理介面（54)，係用以檢查專業詞與 一般詞之誤判，係用以校正問句中的專業詞及一般詞是否 被誤判，如確認為誤判，即予重新定義，並對問句“建構 1226560 式概念腳本”重新進行媒合。 經過前述第-階段的修正處理後， 念腳本”重新進行媒合，並儲存於資料庫中。〃構式概 又前述後端管理機制進一步句紅 組檢查/產生介面（55)，心：有一同義詞或同義詞 、 其主要目的用來觀察誃“读 構式概念腳本”是否有同義字或同 ^ 我組干擾了回應的準 確性，或者真的是新問題；如為新 、 1 ]吨戈為新問題則可依系統訂定的 流程或人工方式產生新的同義字或同義詞組。 由上述可知，本發明主要目的在利用一自然語言解析 技術以解構使用者輸入自然語言之架構，並理解其内容， 進而透過一特別的媒合手段以便在資料庫中才戈到滿足其需 求的資訊内容，利用是項發明將使資訊系統的人機介面更 具親和力，且在資訊的搜尋取得上更臻快速便捷，故本發 明確已具備顯著的實用性與進步性，並符合發明專利要件 ，爰依法提起申請。 【圖式簡單說明】 (一）圖式部分 第一圖··係本發明之系統架構示意圖。 第二圖：係本發明方法之工作流程圖。 第三圖··係本發明“建構式概念腳本，，之内容示意圖 第四圖：係本發明“媒合手段，，之流程圖。 第五圖··係本發明“媒合手段”中邏輯判斷之流程圖 17 1226560 第六圖··係本發明“媒合手段，，之工作示意圖。 第七圖··係本發明“媒合手段，，中邏輯判斷之一工作 示意圖。 第八圖··係本發明“媒合手段，，中邏輯判斷又一工作 示意圖。 選 第九圖··係本發明新產生“建構式概念腳本 ^建構式概念腳本，，間之關係示意圖。Below 40 Below 40: The content of your question may not have been touched by the system. The system will expand new knowledge as soon as possible. Through your questions and interactions, the system will be more intelligent. The following is the system to find the most relevant answer. 〇: 0: 100: Sorry, your question content system may not have been touched ’or your question content has not been clearly explained. Here are some answers that the system may find. From the above description, it can be seen that the information system provided by the present invention using natural language analysis as the core is convenient for users to use natural language to raise their needs for information. The so-called natural language can be text, and 15 1226560 can be Voice, when the voice is spoken, it only needs to be processed by speech recognition technology, and then natural language analysis can be used to achieve the same effect as in the previous embodiment. 0 As the information system should continuously absorb new knowledge to meet the needs of more users, its Means that the information demand put forward by the user is that the system is unable to answer, or the response is not rated well. Also, if the user is not satisfied with the response status, in order to solve this problem, the present invention makes the front-end system step-by-step-back-end management. The 'stage' of the mechanism 'is to provide the following management interface in order to send here the low matchmaking scores and user dissatisfied responses-the mechanism for processing, as shown in Figure 11, which includes: (5 1), for confirming whether the stomach “constructive sketch” according to user needs has been stored in the database and the score Whether it is within the acceptable range, if it is, then no further processing is needed, if not, it is sent to the next management interface; / Regularized management interface (52), which is to simplify the user's question and check the typo. After the normalization, the match will be re-matched. If the match score is acceptable, no other processing is required; a new word check interface (53) is used to check the question after the word break. Whether new words appear, if it is identified as a new word, a new word is added to the database and the part of speech is assigned; a definition correction management interface (54) is used to check the misjudgment of professional words and general words. To correct whether the professional words and general words in the question are misjudged, if it is confirmed to be misjudged, it will be redefined and the question “construct 1226560-style concept script” will be re-matched. After the aforementioned first-stage correction process, the script was re-matched and stored in the database. The structure of the structure and the back-end management mechanism further described the red group check / generation interface (55). Synonyms or synonyms, the main purpose of which is to observe whether "reading the concept script" has synonymous words or homologous ^ Our group interferes with the accuracy of the response, or it is really a new question; if it is new, 1] Ton Gewei For new problems, new synonym or synonym groups can be generated according to the process or manual method set by the system. As can be seen from the above, the main purpose of the present invention is to use a natural language analysis technology to deconstruct the structure of the user's input of natural language and understand it. Content, and then through a special means of matching in order to meet the information content in the database to meet their needs, the use of this invention will make the human-machine interface of the information system more affinity, and in the search for information more refined It is fast and convenient, so the present invention does have significant practicability and progress, and meets the requirements of invention patents, and filed an application in accordance with the law. [1] The first part of the diagram part is the schematic diagram of the system architecture of the present invention. The second diagram is the working flowchart of the method of the present invention. The third diagram is the content of the "constructive concept script of the present invention," The fourth diagram: the flow chart of the "matching means" of the present invention. The fifth diagram ... the flow chart of the logical judgment in the "matching means" of the present invention 17 1226560 The sixth diagram ... is the "matching means of the present invention" Means, the schematic diagram of the work. The seventh diagram is a schematic diagram of the work of the present invention "matching means, one of the logical judgments. The eighth diagram is a schematic diagram of the present invention" the matchmaking means, one of the logical judgments of the work. Select the ninth figure ... This is a schematic diagram of the relationship between the "constructive concept script" and the "constructive concept script".
第十圖：係本發明之回應方式示意圖。 第十〜 〜圖··係本發明後端管理機制之方塊圖。 (二）元件代表符號 (1 〇 ) “建構式概念腳本”資料庫 (1 1 ) ( 2 1 )自然語言解析手段 (2 0 )使用者介面 (3 〇 )媒合機制Fig. 10 is a schematic diagram of a response method of the present invention. Tenth ~ ~ Figures are block diagrams of the back-end management mechanism of the present invention. (2) Symbols of component representation (10) Database of "constructive concept script" (1 1) (2 1) Natural language analysis means (2 0) User interface (3 0) Matching mechanism
(4〇)邏輯解譯單元 (4 1 )調度程式 (5 1 )確認管理介面 (5 2 )正規化管理介面 (5 3)新詞檢查介面 4)定義校正管理介面 (5 5 )同義詞或同義詞組檢查/產生介面 (7 0 ) (80) “建構式概念腳本” (8 1 ) “關鍵事件” 18 1226560 (82) “事件” ( 8 3 ) ( 8 4 ) “關鍵詞組 ( 8 5 ) ( 8 6 )詞組 (8 0 1 )〜（8 0 η )物件(4〇) Logical Interpretation Unit (4 1) Scheduler (5 1) Confirmation Management Interface (5 2) Normalized Management Interface (5 3) New Word Checking Interface 4) Define Correction Management Interface (5 5) Synonym or Synonym Group Inspection / Generation Interface (7 0) (80) "Constructive Concept Script" (8 1) "Key Events" 18 1226560 (82) "Events" (8 3) (8 4) "Keyword Groups (8 5) ( 8 6) phrases (8 0 1) ~ (8 0 η) objects