TW528966B - Natural language dialog system by computer, method and recording medium - Google Patents

Natural language dialog system by computer, method and recording medium Download PDF

Info

Publication number
TW528966B
TW528966B TW089119245A TW89119245A TW528966B TW 528966 B TW528966 B TW 528966B TW 089119245 A TW089119245 A TW 089119245A TW 89119245 A TW89119245 A TW 89119245A TW 528966 B TW528966 B TW 528966B
Authority
TW
Taiwan
Prior art keywords
business
user
dialogue
computer
sentence
Prior art date
Application number
TW089119245A
Other languages
Chinese (zh)
Inventor
Tadamitsu Ryu
Original Assignee
Cai Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from JP11191311A external-priority patent/JP2001022763A/en
Priority claimed from JP29911899A external-priority patent/JP2001117923A/en
Application filed by Cai Co Ltd filed Critical Cai Co Ltd
Application granted granted Critical
Publication of TW528966B publication Critical patent/TW528966B/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/334Query execution
    • G06F16/3344Query execution using natural language analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2452Query translation
    • G06F16/24522Translation of natural language queries to structured queries

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The natural language dialog system by computer consists of input means 101, analysis means 102, switching means 103, chat mode dialog means 104, general mode dialog means 105, business mode dialog means 106, output means 107 and question back means 108 and answer-to-the-question means 109. It can judge the input sentence of the user's natural language, switch to an already prepared dialog means of different mode and respond, and furthermore, if there is deficient information in the input sentence, question back the user, supplement information, carry out the retrieval process and present necessary information to the user.

Description

528966528966

發明説明 【發明所屬之技術領域】 本發明係有關於一種將以自然語言所表現之輸入句可 使其對話的系統,而使其可利用於各種資訊處理機器之介 面者。 習知技藝 近年來’於資訊檢索系統等之對話型資訊處理裝置, 可使一般使用者能容易操作之依自然語言的輸入及對答的 開發正積極進行。其有有關於使不完全之自然語句輸入在 資訊處理裝置時的處理者(特開平9-679〇),有將含在輸入 句中之私示代名詞予以特定而因應文意作對答者(特開平 9-146972)等。 又,在流通業界乃利用網際網路與顧客作通訊已受到 重視於此,即與顧客作自然性之對話且因應必要而可從 顧客獲得業務上所必要之資%的古 受二貝Λ的方法,系統,乃有所期求 特 之 疋,白知的對話系統係從最初就將對話之目的, 定為檢索要求或著某―魏之執行等,而假定其所決定、 目的,僅理解輸人句之文意而已。人員互相之對話乃從益 如是目的之假定等的狀態予以出發明。 語 ---------------- (¾先閱讀背面之注意事項再填寫本頁) 、可· 本發明之目的在於提供一種可判斷由使用者之自秋 所產生的輸人句,並依照該判斷來切換對話模式作對話 且1將使用者引導㈣務對話而可作接近於人類之自 對話的系統、方法及記錄媒體。 而,本發明之另一目的係在 美供一種可判斷其對話 本纸張尺度適用中國國家標準(CNS) A4規格(2】0X297公着) -4- 528966 A7 五、發明説明 之狀況(其輸入句是否為針對系統所發出之詢問的口… 發話狀況),並依其狀況切換輸入模式,以理解輸入而” 使用者之自然語所產生之輸入句,且,判斷其輪入句之 意且執行適切之作業,同時,能將因應該執行之適切的2 答句送還使用者之可接近於人類對話之系統。 【發明之揭示】 本發明之第-態樣係採用電腦所產生之自然語對話系 統及電腦與使用者之自然語料方法,該電腦係至少具有 可由使用者輸人資訊之輸人裝置,而該系統係包含有, 1) 可由性質各異之至少2種對話模式中選擇其—之對 話模式選擇裝置; 2) 可將輸人句作形態要素解析,並將單字列作為資料 列出之分析裝置;及 3) 可判別前述分析之結果來切換前述對話模式之切換 裝置。 、 該對話模式係至少為自以下模式中所選出之兩種。, 1) 依照事先所設定之基本規則僅輪出答問而不獲得資 料之間讀模式。 2) 判定輸人句之動詞且參照事先所準備之辭典以獲得 ㈣詞之格資訊,而從輸人句抽出適合該格之資料 ’右有不相之資訊的格時,就對使用者輸出詢問 該不足格的疑問句’且對使用者要求作不^資訊輸 入而獲得以該格資訊作為項目之資料的一般模式, 及 本紙張尺度適财__準(α^^(2Κ)Χ297公幻 ------------------------裝—— f請先閲讀背面之:沈意事項辱填寫本頁) — *^τ. 線- 528966[Explanation of the invention] [Technical field to which the invention belongs] The present invention relates to a system that allows an input sentence expressed in natural language to make it dialogue, and makes it available to the interface of various information processing machines. Know-how In recent years, dialog-type information processing devices such as information retrieval systems have been actively developed for natural language input and answering which can be easily operated by general users. There are those who deal with incomplete natural sentences when they are input into the information processing device (Japanese Patent Application Laid-Open No. Hei 9-679), and those who specify private pronouns included in the input sentence and respond to the meaning of the text (specially) Kaiping 9-146972) and so on. In the distribution industry, the Internet has been used to communicate with customers, which means that they have a natural dialogue with customers and can obtain the necessary capital for business from customers if necessary. Method and system are expected to be special. Baizhi's dialogue system has set the purpose of the dialogue as a retrieval request or the execution of a certain Wei from the beginning, assuming its decision and purpose, only understanding Just the meaning of the input sentence. The dialogue between people is invented from the state of benefit, such as the assumption of purpose. Language ---------------- (¾Read the precautions on the back before filling this page), but · The purpose of the present invention is to provide a method that can be judged Enter a sentence, and switch the dialogue mode for dialogue in accordance with the judgment. The system, method, and recording medium can be used to guide the user to conduct a dialogue and be close to human self-talk. However, another object of the present invention is to provide a paper in the United States that can judge its dialogue. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (2) 0X297. -4- 528966 A7 V. Status of the invention description (the Whether the input sentence is a response to a query issued by the system ... Speaking status), and switch the input mode according to the situation to understand the input and the input sentence generated by the user ’s natural language, and determine the meaning of the turn sentence It can also perform appropriate tasks, and at the same time, it can return the appropriate 2 answers that should be executed to the user's system that is close to human dialogue. [Disclosure of the Invention] The first aspect of the present invention uses computer-generated nature. Language dialogue system and natural corpus method of computer and user, the computer has at least input device capable of inputting information by the user, and the system includes, 1) at least two dialogue modes with different properties Choose the device for dialogue mode selection; 2) An analysis device that can analyze input sentences as morphological elements and list single words as data; and 3) can judge the results of the foregoing analysis Change the switching device of the previous dialogue mode. The dialogue mode is at least two selected from the following modes., 1) According to the basic rules set in advance, only the question-and-answer mode is rotated and no data reading mode is obtained. 2 ) Determine the verb of the input sentence and refer to the dictionary prepared in advance to obtain the case information of the epithet, and extract the data suitable for the case from the input sentence. The inadequate interrogative sentence 'and the general pattern of obtaining information that uses the case information as an item without user information input, and the paper scale is suitable for wealth __ 准 (α ^^ (2Κ) × 297 公 幻- ----------------------- Equipment—— f Please read the following first: Concerns and Disgraceful Fill in this page) — * ^ τ. 线-528966

、發明說明 3)將事先所设定之業務項目作為基礎並從輸入句獲得 對該業務項目相稱之資訊,若有未填補之項目時, 就對使用者輸出詢問該不足項目的疑問句並對使用 者作不足資訊輪入要求,而獲得對應於該不足業務 項目之資料的業務模式。 或者,在至少具有可由使时輸入資訊之輸入裝置之 電腦所產生的自然語對話系統上,亦能作成包含具備因應 系統管理者之業務的必要而參照事先所設定之業務用項目 ,且從輸入句獲得合適於該業務用項目之資訊的裝置,與 有未填補之項目時,將詢問該不足項目之疑問句對使用者 輪出之裝置,以及對使用者作不足資訊輸入要求而獲得對 應於未填補之業務項目的資料之裝置的業務模式對話裝置 之電腦所產生之自然語對話系統及電腦與使用者之自然語 對話方法。於此,業務模式亦能再分割為業務用項目相異 之多數的業務模式。 收 ----------------,…: (請先閱讀背面之注念事項再填寫本頁) ’訂丨 又,在於使用者之輸入句的分析上乃具備由於參照開 出之貧料與儲存該資料所屬之項目的資料庫,而判定應μ 納之業務項目的順序不同處理裝置,同時,可接受遍及多 數業務用項目之資料而事先不指^輸人項目亦可將該輸入 貧科填補於不足業務用項目,且生成對象資料,即使對於 二先之’.D構不太理解之使用者,以不同順序輸入資料亦能 柔軟對應。而且’存取在對該業務所關連之資料庫並將該 貝枓庫之綱目作為業務用項目而獲得,就能使對該業務有 關之系統管理者削減要一一設定應生成目的資料之綱目的 本紙fe:尺度翻巾關_準(⑽)A顿格(2獻公 -6- 528966 五、發明説明(4 ) 費時費事工夫。 又,於滿足從事先所設定之時.間的經過, 數的對話所選擇的至少為其-之條件的區間,未進 仃業務用對話時,若將包含執 誘導穿晉作幻塞m 所設定之質問的業務 =乍輕成要素,乃對市場話動上有效。並且,於 則述業務模式之對話上,若以未用於 有有因其:模式啟《,即記憶前述未填補伽^ 〜义要而询問不足項目之項目監視裝置作為 ,就能對使用者繼續詢問所必要之資料至最後。要素 於本發明之第2態樣係採用含有: υ判斷對話之狀況來切換輸人模式之模式切換裝置; 2)對輸人句作形態要素分析 ⑽前述形態要素分析之結果作構句分:置且生成附有 表不構句上之格的標籤之單字列表的作用分析襄置 4) 輸入對於該系統所輸出之疑問句的回答句而即判斷 時,k則述形態要素分析之結果列出品詞為名詞之 單字且抽出作為回答之反問補充裝置; 5) 可於判疋輸入句後’從將前述構句表現資料作為知 識而儲存在資料庫之登錄處理、檢索已儲存在前述 賁料庫之知識之檢索處理、催促使用者輸入構句不 完全之輸入句不足格之反問處理以及使預先設定之 功能啟動之功能執行處理等選擇至少一輸入句處理 之輸入句判定裝置; 本纸張尺度賴巾標準(CNS) M祕(2獻297公3. Description of the invention 3) Based on the business items set in advance as the basis and obtain information commensurate with the business items from the input sentence, if there are unfilled items, the user will output a question asking the insufficient item and use it. Those who make insufficient information rotation request and obtain the business model corresponding to the insufficient business items. Alternatively, a natural language dialogue system generated by a computer having at least an input device capable of inputting information can also be prepared to include items for business that are set in advance according to the needs of the system manager's business, and input from the input A device that obtains information suitable for the business use item, and an unfilled item, will ask the question of the insufficient item to the device that the user turns out, and make a request for insufficient information input to the user to obtain the corresponding item. The business model of the device to fill the business item information. The natural language dialogue system generated by the computer of the device and the natural language dialogue method between the computer and the user. Here, the business model can also be divided into business models with a large number of different business items. Receive ----------------, ...: (Please read the notes on the back before filling out this page) 'Order 丨 Also, it is based on the analysis of the user's input sentence. The order of business items that should be accepted should be determined by referring to the developed poor materials and the database that stores the items to which the data belongs. At the same time, the data for most business items can be accepted without prior input. The project can also fill the input poor section with insufficient business items and generate the target data. Even for users who do not understand the structure of Erxian '.D, input data in different orders can be softly corresponded. In addition, 'access to a database related to the business and obtaining the outline of the Belarus library as a business project can reduce the system administrators related to the business to set the outline of the data to be generated one by one的 本 纸 fe: Scale turning towel off _ quasi (⑽) A Dunge (2 Xian Gong-6- 528966 V. Description of the invention (4) Time-consuming and labor-intensive. In addition, to meet the time from the time set in advance. The interval selected by the number of dialogues is at least a condition of-. If the business dialogue is not used, if the business including the challenge set by the persuasion m is considered as the element, it is the market The speech is effective. And, in the dialogue of the business model, if it is not used for a reason: the mode is activated, that is, the project monitoring device that memorizes the above unfilled gamma ^ ~ meaning and asks for insufficient items as , The user can continue to inquire the necessary information to the end. The elements in the second aspect of the present invention are: a mode switching device that includes: υ judges the state of the dialogue to switch the input mode; 2) forms the input sentence Element analysis The results of the prime analysis are used to construct the sentence: analysis of the role of placing and generating a list of words with labels on the unstructured sentence. 4) When entering the answer sentence of the question sentence output by the system, that is, judgment, k The result of the analysis of the morphological elements lists the word with the noun as the noun and extracts it as a supplementary device for rhetorical questions; 5) After judging the input sentence, 'from the foregoing sentence formation performance data as knowledge and stored in the database for registration processing Selecting at least one input sentence to process the search process of retrieving the knowledge that has been stored in the aforementioned database, urging the user to input incomplete sentence formation, inadequate rhetorical question processing, and the function execution process to activate a preset function. Input sentence judging device; The paper standard Lai Jin Standard (CNS) M Secret

----. (請先閲讀背面之注意事項再填寫本頁) -訂— :線丨 -7- 528966----. (Please read the notes on the back before filling out this page) -Order —: Line 丨 -7- 528966

)依照前述輸入句判定而進行輸入句處理的輸入句肩 理裝置;及 ----------------f:; Γ,?先閲讀背面之注念事項再填寫本頁) 取前述輸人句處理裝置之結果q成輸出句並輸之 對答裝置; 作為特徵之電腦所產生的自然語對話系統。 而於忒系統所執行之方法含有以下步驟: 1) 判斷對話之狀況來切換接受輸入之模式; 2) 依前述模式解析輸人句後,再構成為由附有解析結 果標籤之單字所構成之列表構造之資料(以下稱「知 識化資料」)並理解該輸入句。 3) 分析前述知識化資料且判斷由知識之傳授、知識之 .、訂— 確認、知識之檢索以及預先設定之功能動作的依賴 而所選出之至少其一文意; 4) 執行對應前述判斷結果之作業;及 5) 生成對應前述作業執行之回答句後,輸出並回答。 於有關上述第2態樣之發明上,輸人模式至少乃由通常 :式與對於該系統所輸出之疑問句的回答句有輸入時的回 答模式之2種所構成。 輸入句乃至少有由鍵盤輸入之正文句與結由聲音認識 裝置由正文化之文章所選之中的!種。對於對話之狀況乃以 在瞬前該系統是否將疑問句輸出予該對話之對象就能作判 斷0 於上述電腦與使用者之自然語對話方法上的標記乃可 參照由事先所作成之辭典與從將輸人句作構句解析之 娜尺度賴^格(2WX297J^ 528966 A7 B7 五、發明說明 Φ Φ 所選之中的至少之1種而予進行。 【本發明之實施態樣】 以下,參照圖面針對本發明之實施態樣作說明。第i 圖為本發明之實施態樣的方塊圖。於該方法,系統乃讀入 儲存在記憶媒體之程式,並依使其動作由該程式所控制之 電腦而實現。 於本實施態樣之系統,係由輸入裝置101、分析裝置102 、切換裝置103、閒讀模式對話裝置1〇4、一般模式對話裝 置1〇5、業務模式對話裝置1〇6、輸出裝置1〇7、反問裝置⑽ 及反問回答裝置109所構成。 輸入裝置101係將由使用者之自然語所發聲之輸入 作聲音認識且作正文變換而輸入至分析裝置⑽。於此, ,入裝置101,乃採用聲音認識裝置,惟,輸入裝置並不 定於此,而使用者以鍵盤直接打人輸人句,或將所寫入 文章以OCR認識裝置讀取均可以。 分析裝置102係對輸入句進行形態要素解析,而生成有 附與品詞資訊之單字列並傳送至切換裝置i q 3。切換裝置乃 判別輸人句而執㈣讀模式對話裝置104,-般模式對話裳 置105或著’業務模式對話裝置! Q6等性質相異之對話模 式的切換。輸出裝置107乃利用習知技術之聲音合成裝置 將於對話裝置所生成之回答句作聲音輸出。於對話裝置右 有不足資料之時就因應必要而使反問裝置1〇8起動並對使 用者執行詢問之輸出。反問後從使用者之輸入句乃於反問 回答裝置接受且補完不足資料而進行對活。於此,在輸出 句 在 限 之 而 若 (請先閲讀背面之注念事項再填寫本頁) .裝-) Input sentence processing device for input sentence processing according to the aforementioned input sentence judgment; and ---------------- f :; Γ ,? Read the notes on the back before filling (This page) Take the result q of the input sentence processing device into an output sentence and input the answer device; a natural language dialogue system generated by a computer as a feature. The method executed by the Yu system includes the following steps: 1) Judging the status of the dialogue to switch the mode of accepting input; 2) After analyzing the input sentence according to the foregoing mode, it is constituted by a word with a label of the analysis result List construction data (hereinafter referred to as "informative data") and understand the input sentence. 3) Analyze the aforesaid knowledgeable data and judge at least one of the meanings selected by the transfer of knowledge, knowledge., Order-confirmation, knowledge retrieval and pre-defined functional actions; 4) execute the corresponding judgment results Assignment; and 5) After generating a response sentence corresponding to the execution of the aforementioned assignment, output and answer. In the invention of the second aspect described above, the input mode is composed of at least two types: a normal mode and a response mode when an answer sentence to the question sentence output by the system is input. The input sentence is at least a text sentence input by the keyboard and a sound recognition device selected by the article of Zhengwen! Species. The state of the dialogue can be judged by whether the system outputs the interrogative sentence to the object of the dialogue before the moment. 0 The marks on the natural language dialogue method of the computer and the user can refer to the dictionary and the pre-made dictionary. The scale of the analysis of the input sentence as a sentence analysis Lai ^ lattice (2WX297J ^ 528966 A7 B7 V. Description of the invention Φ Φ at least one of the selected and carried out. [Implementation of the present invention] The following, with reference to the drawings The embodiment of the present invention is described below. Figure i is a block diagram of the embodiment of the present invention. In this method, the system reads a program stored in a memory medium and controls its actions according to the program. The system in this embodiment is composed of input device 101, analysis device 102, switching device 103, idle reading mode dialog device 104, normal mode dialog device 105, and business mode dialog device 10. 6. The output device 107, the questioning device ⑽, and the questioning and answering device 109 are composed of the input device 101, which recognizes the input made by the user's natural language as sound recognition and converts the text to the input. Analysis device ,. Here, the input device 101 uses a voice recognition device, but the input device is not determined here, and the user directly enters the input sentence by typing on the keyboard, or reads the written article with the OCR recognition device. Any analysis is possible. The analysis device 102 analyzes the morphological elements of the input sentence, generates a single word string with the word information and sends it to the switching device iq 3. The switching device judges the input sentence and executes the reading mode dialogue device 104, -Normal mode dialogue 105 or "business mode dialogue device!" Q6 and other dialogue modes with different characteristics. The output device 107 is a sound synthesis device using conventional technology to output the answer sentence generated by the dialogue device as a sound. When there is insufficient data on the right side of the dialog device, the questioning device 108 is activated as necessary and the user outputs the query. After the questioning, the input sentence from the user is accepted by the questioning answering device and the insufficient information is filled to complete the matching. At this point, the output sentence is limited (please read the notes on the back before filling out this page).

-線I 本紙張尺度朝巾關緖準(⑽A4規格(2淑297公 -9- 528966-Line I The size of this paper is toward Toki Sho (准 A4 size (2 ladies 297 males -9- 528966

五、發明説明 先 此 不 置 乃使用聲音合成裳置,惟,能表現畫面顯示等語言者,使 用任何裝置均無問題。 切換裝置103係查對輸入之單字列且參照事先所準備 之規則庫檢測出問㉟,招呼、詞不達意、時間之疑問,即 判定為閒談模式,且與「招呼」、「詞不達意」等之判別結 果起而將除去助詞,語尾之指示表明語(名詞、形容詞、 動判)作為參數,而傳送至閒談模式對話裝置。 第2圖為所參照之規則庫之一例。例如,「好熱呢丨」 輸入句時,即,與「詞不達意」之判別結果而使「熱」 之語幹的參數傳送至閒談模式對話裝 置104。又,「現在幾 點」時,使「時間之質問」的判別結果與「現在」的參數 一起傳送至閒談模式對話裝置104。 在切換裝置103於所傳送來之單字列檢測出業務之關 鍵字時,乃將從關鍵字所判別之「業務名」傳送至業務模 式對話裝置106而起動業務才莫式對話,業務之關鍵字應事 準備業務用關鍵字列表而於關鍵字檢測時作參照為宜。& 時,設定多數之業務而準備各業務之關鍵字列表,乃對多 數之業務就可作對應,是故,更佳,第3圖為業務之關鍵字 •業務項目之列表的一例。若對規則庫或關鍵字列表均 符口日守,則,所分析之單字列就傳送至一般模式對話裝 105 〇 ^ 第4圖係詳細表示閒談模式對話裝置1〇4者,乃由規格 選擇裝置40卜規格基礎402、對答生成裝置4〇3、計時器4〇4 所構成。 紙張尺度相中關家標準(⑽)Μ規格⑵⑽撕公 t ..... f 先閱讀背面之';iT&事項再場寫本頁) -10-V. Description of the Invention Firstly, this is not to use sound synthesis, but those who can express languages such as screen display can use any device without any problem. The switching device 103 is to check the input word string and to refer to the rule library prepared in advance to detect questions, greetings, unsatisfactory words, and time questions, that is, it is determined to be a chat mode, and is related to "hello" and "words unsatisfactory" As a result of the discrimination, the auxiliary words are removed, and the indications at the end of the word (noun, adjective, verb) are used as parameters, and they are transmitted to the chat mode dialogue device. Fig. 2 is an example of the referenced library. For example, when the sentence "Is hot" is inputted, that is, the parameter of "Hot" is transmitted to the chat mode dialogue device 104 with the result of the discrimination with the word "unsatisfactory". At the time of "what time", the determination result of "question of time" is transmitted to the chat mode dialogue device 104 together with the parameter of "now". When the switching device 103 detects the keyword of the business in the transmitted single-word string, the "business name" identified from the keyword is transmitted to the business model dialog device 106 to start the business mode dialogue. It is advisable to prepare a business keyword list and use it as a reference during keyword detection. & In the case of setting up a large number of businesses and preparing a keyword list for each business, it is possible to correspond to most businesses. Therefore, it is better. Figure 3 is an example of a business keyword list. If both the rule base and the keyword list are in line, the analyzed words will be transmitted to the general mode dialogue device 105. Figure 4 shows the chat mode dialogue device 104 in detail, which is selected by the specifications. The device 40 includes a specification base 402, a answer generation device 403, and a timer 404. Guan Jia standard (⑽) M specifications ⑵⑽ tearing t ..... f first read '; iT & Matters and write this page on the back) -10-

X 就 列 之 528966 五、發明説明 〇 從切換裝置所送出之判別結果與參數資訊就輸入至規 格選擇裝置4G1。於前例,「好熱呢!」之輸人句,其判別 結果為「表達不完善」、參數為「熱」。若為「現在幾點」 ,其判別結果之「時間之質問」,參數資訊之「現在」就輸 入至規格選擇裝置401。 規格選擇裝置401乃參照基本規格4〇2,而將輸出句生 成資訊傳送至應答生成裝置403。第2圖為規則庫之一例。 在對答生成裝置403 ’若輸入為「好熱呢丨」之時,就接受 於第2圖所示之輸出二「是的」之規格,且以其原狀就可生 成對答。若輸入|「現在幾點」之時,就從計時器綱取得 對應於參數值「現在」之輸入時的時間,而生成「⑼點 X分」的對答綱目。於對答生成裝置403所生成之回答句 傳送至輸出裝置107。 第5圖係詳細表示一般模式對話裝置1〇5者,其由構句 資訊生成裝置501、動詞格列表資料庫5〇2、疑問句判別裝 置503、檢索裝置504、一般知識資料庫5〇5、知識生成裝置 5〇6、隨聲附和生成裝置5〇7、以及回答生成裝置⑽,㈣ 成。 構句資訊生成裝置501乃從由輸入句所分析之單字 貝料抽出動詞,且由動詞之格列表資料庫5〇2呼出該動詞〜 格列表,並生成由適用於格資訊及其格之單字資料所㈣ 之目標(以下’稱為「構句資料」)。此時,於所呼出之格 列表之中無法依單字資料填補之格,即,檢測不足格之不 足格資訊就傳送至回問裝置108。回問裝置1〇8乃將所受取 本紙張標準(⑽)A4規格(21GX297公釐)— ^------------------、盯------------------緣 (枝?先閲讀背m;之注意事项再填寫本頁) -11 - 主 528966 五、發明説明(< 不足格貝Λ ’且參照未圖示之反問輸出句樣規生成對使用 者反問之文章並輸出,而催促從使用者之回答輸入。例如 才右為主格之時’就以「那-位? j作反問,而不 ^若為時制,即生成「何時?」之反問且輸出。有反問 句輸出之時,從使用者之輸入句就直接輸入於反問回答裝 置9且攸回合之輸入句抽出而填補不足格之單字資料。 不足格均被填補之後,前述所生成之構句資訊乃於疑 問^判別裝置5〇3進行疑問句或肯定句之判定。若為疑問句 次2檢索裝置5〇4而將該構句資訊作為鍵,乃對一般知識 ^ 5〇5進'目標檢索,並將命中之目標傳送至回答生成 08。在回答生成裝置,對於有疑問詞之疑問句,乃將 相當於疑問詞之格的單字資料從成功之目標開出,且參照 θ之回D句樣規生成回答句。若為查詢無疑問詞之輸 〇句的真偽之輸人句之時1有目標檢索之結果成功的目 才示之時,就判定為「真」,而為未成功之時,即判定為「偽 」’亚依照未圖示之回答句樣規’若為真就生成「就是那樣 」之回答句’而為偽即生成「 , q 至輪出裝置1〇7。 疋」回答句共傳送 、,/、對於有疑問詞之疑問句的回答生成,卩「吉田何時去 〜合?」之疑問句為例稍作詳細說明。 該疑問句於構句資訊生成裝置501乃成為{動作(去)、 格(吉田)、目的地(涉谷)、時間(何時)}之構成資訊(目標 口檢索裳置504乃將{動作(去)、主格(吉田)、目的地(涉谷 ;作為鍵而檢索一般知識資料庫505。若無命中之目標時, 紙張尺度適用中國國家標準(CNS) Α4規格(210X297公楚)X is listed in 528966 V. Description of the invention 〇 The judgment result and parameter information sent from the switching device are input to the specification selection device 4G1. In the previous example, the input sentence of "It's hot!" Was judged as "incomplete expression" and the parameter was "hot". If it is "what time is it", the result of the judgment is "question of time", and the "now" of the parameter information is input to the specification selection device 401. The specification selection device 401 refers to the basic specification 402, and transmits the output sentence generation information to the response generation device 403. Figure 2 shows an example of a rule base. When the answer generation device 403 ′ is “I ’m so hot”, it accepts the specification of output two “Yes” as shown in FIG. 2, and generates the answer in its original state. When inputting "" What is the current time ", the time corresponding to the input value of the parameter value" Now "is obtained from the timer outline, and the answer outline of" point X points "is generated. The answer sentence generated by the answer generation device 403 is transmitted to the output device 107. FIG. 5 shows the general mode dialog device 105 in detail, which is composed of a sentence formation information generating device 501, a verb case list database 502, an interrogative sentence discrimination device 503, a retrieval device 504, and a general knowledge database 505, The knowledge generation device 506, the accompanying sound generation device 507, and the answer generation device ⑽ are completed. The sentence formation information generating device 501 extracts a verb from the vocabulary material analyzed by the input sentence, and calls out the verb ~ case list from the verb's case list database 502, and generates a word suitable for the case information and its case. The purpose of the data (hereinafter referred to as "sentence data"). At this time, in the list of called cells, the cells that cannot be filled with the single-word data, that is, the insufficient information of the detected insufficient cells are transmitted to the questioning device 108. The questioning device 108 is to receive the standard (⑽) A4 specification (21GX297 mm) of this paper — ^ ------------------, stare ----- ------------- Fate (branches? Read the precautions before filling in this page, and then fill out this page) -11-Master 528966 V. Description of the invention (< Insufficient Ge Bei Λ 'and refer to The rhetorical output sentence pattern shown in the figure generates and outputs the rhetorical question to the user, and urges input from the user's answer. For example, when the right is the main case, 'the-digit? J is asked instead of ^ If it is time-based, the question and output of "when?" Is generated. When the question is output, the user's input sentence is directly input to the question answering device 9 and the input sentence of the round is extracted to fill the insufficient word data. After all the deficiencies are filled, the generated sentence information is judged by the question ^ discriminating device 503 as a question or affirmative sentence. If it is the question sentence 2 and the retrieval device 504 uses the sentence information as the key, It is based on general knowledge ^ 505 and searches for the target, and transmits the hit target to the answer generator 08. In the answer generator, for question words The interrogative sentence is the one-word data corresponding to the case of the interrogative word from the goal of success, and the answer sentence is generated by referring to the pattern of the D sentence of θ. At that time, when the result of the target search is successful, it will be judged as "true", and if it is unsuccessful, it will be judged as "false". The answer sentence "that is" is generated when it is true, and ", q to the turn-out device 1007 is generated if it is false. The answer sentence is transmitted in total, /, and the answer sentence for the question sentence with question words is generated," Yoshida When will I go ~ come together? ”As an example, I will explain it in detail. The question sentence in the sentence formation information generating device 501 becomes {action (go), grid (Yoshida), destination (Shibuya), time (when)}. The composition information (target mouth retrieval clothes set 504 is to retrieve {action (go), main grid (Yoshida), destination (Shibuya; as the key and search the general knowledge database 505. If there is no hit target, the paper scale applies to the Chinese country Standard (CNS) Α4 specification (210X297 Chu)

---- (-*_?先閲讀背面之汶*事項再塡寫本頁) .^τ— 0, -12- 528966 五、發明説明(ι〇 ) 就參照未圖示之回答句樣規而生成「對其不清楚」之回答 句且傳送至輸出裝置107。 先 閲 讀 背 面 之 注 ,若目標命中時,而其為(動作(去),主格(吉田)、目的 也(二谷)時間(昨天》之目標之時,即參照未圖示之回答 句稱規「其為(回答)」且將附與符合疑問詞之格的時間: 之單字,即,將「昨天」加入於(回答)而生成「其為昨日 」之回答句且傳送至輸出裝置107。 訂 第6圖係詳細表示業務模式對話裝置1()6者,其由業務 模式監視裝置6(H、模式遷移裝置6〇2、業務特定裝置6〇3 、項目監視裝置604、檢索條件之收集裝置6〇5,格列表之 順序不同處理裝置606、檢索條件緩衝器6〇7、檢索/回答句 生成裝置608、格列表緩衝器_、格列表抽出裝置61〇、以 及業務知識資料庫611,所構成。 ,從切換裝置103所選出之業務關鍵字以及參數資訊就 傳运至業務特定裝置6〇3。業務模式亦有因應必要而依由業 務模式監視裝置之指示而作特定。所特定之業務模式就傳 达至項目監視裝置6G4,並依從業務之關鍵字、業務用項目 的列表所抽出之業務用項目而整理參數資訊,且作為業務 資訊目標生成。此時’對應於該項目列表之某項目的參數 貝Λ有所不足且由☆參照才各列表緩衝器6〇9而依格列表抽 出裝置610所檢測出之時’乃對反問裳置刚傳送該不足項 目。而後,就生成催促輸入相當於不足格之資訊的反問句 且輸出至使用者。從使用者之回答乃經由反問回答裝置丨〇8 而收集在檢索條件之收集裝置605。 本紙張尺度適用中國國家標準(CNS) Α4規格(2Κ)><297公釐) -13- 528966---- (-* _? First read the Wen on the back * and then write this page). ^ Τ— 0, -12- 528966 V. Description of the invention (ι〇) Refer to the sample sentence rule that is not shown in the figure The answer sentence "Unclear" is generated and transmitted to the output device 107. First read the note on the back. If the target is hit, it is (action (go), main grid (Yoshida), goal also (Yutani) time (yesterday) when the target is, then refer to the answer sentence not shown. "It's (Answer)" and the time when it will be attached to the grid that matches the question: that is, adding "Yesterday" to (Answer) to generate a "It's Yesterday" answer sentence and send it to the output device 107. Figure 6 shows the business mode dialog device 1 () 6 in detail, which is collected by the business mode monitoring device 6 (H, mode migration device 602, business specific device 603, project monitoring device 604, and search conditions). Device 605, the processing order of the cell list is different, the processing device 606, the search condition buffer 609, the search / response sentence generating device 608, the cell list buffer_, the cell list extraction device 61, and the business knowledge database 611, The business keywords and parameter information selected from the switching device 103 are transmitted to the business specific device 603. The business model is also specified according to the instructions of the business model monitoring device as necessary. The determined business model is transmitted to the project monitoring device 6G4, and the parameter information is organized according to the business keywords extracted from the business keywords and business items list, and generated as business information targets. At this time, 'corresponds to the item list' When the parameter of an item is insufficient and it is detected by the ☆ reference list buffer 609 and the in-grid list extraction device 610, the question is just transmitted to the questioner. Then, it is generated. Urgently input rhetorical questions equivalent to inadequate information and output them to the user. The answers from the users are collected in the search condition collection device 605 via the rhetorical question answering device 丨 〇8. This paper standard applies the Chinese National Standard (CNS) Α4 size (2Κ) > < 297 mm) -13- 528966

::?先閲讀背面之注意事項再填寫本頁) 從使用者之回答為未適樣於剛才之反問所要求之格的 貧訊時’只要其適稱於現在起動中之格列表之不足格,即 由格列表之順序不同處理裝置606附與所適合之格資訊而 補完格列表。滿足格列表之一切的目標就暫時保存於檢索 條件緩衝器607,且於檢索/回答句生成裝置6〇8而將業務知 識資料庫611作目標檢索並從所命中之業務知識生成回答 。所生成之回答就傳送至輸出裝置1〇7且使用聲音合成裝置 、監控器而對使用者作顯示。 再以「想買紅色的車」之輸入句為例作說明。輸入於 輸入裝置HH之例句,乃在分析裝置11〇2分析為{形容詞( 、工)m尾(l、)(i)、名巧(車)、助詞(会)(〇)、動詞(買)、語尾 訂· (H L、)(i Ta i)}。該分析裝置在切換裝置1〇3上乃判定為 業務用模式,而使{業務料(車、f)、參數(紅)丨送出至業 務特疋裝置603。業務特定裝置就參照業務之關鍵字,業務 項目列表(第3圖),且將業務特定為「中古車販賣」,乃在 項目監視裝置而將「中^車販賣」的特定業務傳送至項目 &視裝置604。 項目列表抽出裝置61G乃參照在第3圖所示之業務的關 鍵字’業務用項目列表而取得(廠商、價格、任務、顏色) 之項目列表。於此,於項目列表巾「顏色」以「紅」之單 字資料就可,惟其他項目艿士、& τ 口 π 、乃成為不足項目。項目監視裝置 604就將不足項目傳送至^ ^ ^ ^ ^ 「J攻置108。反問裝置1〇8乃依照 未圖示之反問句樣規而輪出「 —广士 0 那一家廠商?」的反問句。 於此’從使用者之回答A「 ”、、日產」而相稱於詢問項目時, 本紙張尺度適用巾關家鮮(⑽) -14- '12 528966 五、發明説明 就以此補完不足項目。 例如為「5人座之手妞由土 車較好」而對反問句不相稱之回 答輸入時,乃在項目列表 冉之口 在成為不足項目之(廠商、:ΠΓ裝置606上亦與現 於「任務」之項目的「手排、務)相照合,取得符合 作反覆’且以{«(日產)、價格_萬 。 、顏色(紅之情況使資料均收集於業務用項(手排) 索/回答生«置、寺,於檢 知識乃以影像資料或動書資 古車的 輸出裝論)。 -貝科而輪出至使用者之監控器( 其次,針對依能判斷對話之狀況的 系統及方法,參照圖面說明。 目一對活 第7圖係湘能卿本發明之第2g樣有 況的電腦之自然語對話方法與系統之方塊圖。"之狀 漆 本實施態樣上之系統’係由模式判斷裳置1U、單字八 解:作用分析裝置112、單字分解一名詞抽出裝置〗心 入^定裝置U4、格列表確認裝置115、反問 二 、輸入句處理裝置117、以及應答裝置m,所構成。置叫 …模式判斷裝置⑴’係收到輸入句處理敦置… 後就切換輸人模式。預設下為通f模式,且從 入裝置(鍵盤、聲音認識裝置)之輸人句就傳送於單 :::1=置112。於輸入句處理裝置117進行詢問:理 有/、通知之時,乃成為回答模式並對單字 抽出裝置113傳送輸入句。 名㈣ 本紙張尺度適用中國國家標準(CNS) A4規格(210X297公釐) -15- 〜〇夕0〇 A7 ~____Β7__ 五、發明說明(Η ) ' "~' ~- 》單字分解、作用分析裝置112乃將輪入句件形態要素分 彳’而且進行構句解析。並參照未圖示之構句法則及動詞 丄:格列表資料庫’而生成由有附與構句資訊(以下,稱為「 ^ ‘25作用」)之單字的列表所成之目標。 :五例如,「Α之Β為C」之輸入句,從構句法則乃成為(主 忐之:用⑻、資料(八之…:補語之作用(無)、資料(〇:} ‘又,D將去E」時,將參照所檢出之動詞(去)的格 列表而生成{動詞(去)、作用((i )(wa)、資料⑼、作用⑹⑽ 、資料(E)}。 單字分解一名詞抽出裝置113於形態要素分析之結果 從附與品詞資訊而開出之單字抽出名詞之單字,並將 該單字送出至反問補充裝置116。 輸入句判定裝置114就檢測出句尾之「?」「妒」(ka)及 疑問屑之存在且分別分開使肯定句至格列表確認裝置Η $ ❿疑問句至輸入句處理裝置117且送出輸入句目標。 次袼列表確認裝置115乃從所檢出之動詞而由未圖示之 貧料庫啤出格列表,且與輸入句之單字的構句資訊相照合 ,若有未填補之格(不足格)之時,就將不足格資訊送出至 反問補充裝置11 6。若無不足格之時,輸入句目標就傳送至 輸入句處理裝置117。 反問補充裝置116將參照表示其詳細之第8圖作說明。 解析從通常模式所接受之輸入句的結果,若檢測出有不足 秸之時,就彳之格列表確認裝置傳來不足格資訊。不足格暫 捋保持裝置211乃將不足格資訊儲存在緩衝器212之後,就 本紙張尺度適用中國國家標準(CNS) Α4規格(2】〇X297ij^-::? Read the precautions on the back before filling this page.) When the user ’s answer is that the poor information is not suitable for the box requested in the previous rhetorical question, as long as it is considered to be inadequate for the list of boxes currently being activated. That is, the grid list is completed by attaching the appropriate grid information to the processing device 606 in a different order of the grid list. The objects satisfying all the grid lists are temporarily stored in the search condition buffer 607, and the business knowledge database 611 is searched for by the search / response sentence generating means 608, and a response is generated from the business knowledge hit. The generated answer is transmitted to the output device 107 and displayed to the user using a voice synthesis device and a monitor. Take the input sentence of "want to buy a red car" as an example. The example sentence entered in the input device HH is analyzed by the analysis device 1102 as {adjective (, work) m tail (l,) (i), name Qiao (car), auxiliary (hui) (〇), verb (buy ), Suffix · (HL,) (i Ta i)}. This analysis device judges that it is a business mode on the switching device 103, and sends {business data (car, f) and parameters (red)) to the business special device 603. The business specific device refers to the keywords of the business, the business item list (Figure 3), and specifies the business as "medium car sales". The project monitoring device transmits the specific business of "medium car sales" to the project & amp Depending on device 604. The item list extraction device 61G is an item list (manufacturer, price, task, color) which is obtained by referring to the business item list of business keywords' shown in FIG. 3 (maker, price, task, color). Here, the word data of "red" in the item list can be used, but the other items, "& τ mouth π", are insufficient. The project monitoring device 604 transmits the insufficient items to ^ ^ ^ ^ ^ "J attack set 108. The questioning device 108 turns out" —Which company is Guangshi 0? " Rhetorical questions. Here's the answer from the user's answer "A", "Nissan", which is commensurate with the inquiry item. This paper size is suitable for towels. (14) -14- '12 528966 V. Description of the invention This completes the insufficient items. For example, when the input of the answer to the rhetorical question is "the five-seater hand girl is better from a dirt car", it is in the project list that it has become an insufficient project (manufacturer, ΠΓ device 606 is also the same as now "Task" items of the "hand platoon, work" match, get consistent and repeated "and use {« (Nissan), price _ 10,000., Color (in the case of red, all data are collected in business items (hand platoon)) Asking / responding students «set, temple, knowledge of inspection is based on the image data or the output of ancient books of ancient books).-Beko turns to the user's monitor (Second, the status of dialogue can be judged according to the ability to judge The system and method are described with reference to the drawings. Figure 7 is a block diagram of the natural language dialogue method and system of the 2g-like computer of the present invention in Xiang Nengqing's invention. The system on the sample is a model judgement 1U, eight-character solution: role analysis device 112, single-word decomposition-noun extraction device〗 Heart-setting device U4, cell list confirmation device 115, question two, input sentence processing device 117 And the answering device m. Place ... The mode judging device ⑴ 'is to switch the input mode after receiving the input sentence processing and resetting ... The default mode is f mode, and the input sentences from the input device (keyboard, voice recognition device) are transmitted to the ticket :: : 1 = Set to 112. When the input sentence processing device 117 makes an inquiry: When the notification is accepted, it becomes the answer mode and transmits the input sentence to the single-word extraction device 113. Name ㈣ This paper standard applies Chinese National Standard (CNS) A4 Specifications (210X297 mm) -15- 〜〇 夕 0〇A7 ~ ____ Β7__ 5. Description of the invention (Η) '" ~' ~-"The word decomposition and role analysis device 112 divides the form elements of the sentence into turns." Sentence analysis is also performed. With reference to unillustrated construction rules and verbs 丄: Case List Database ', a list of words with attached sentence formation information (hereinafter referred to as "^ '25 role") is generated. The goal of success: Five. For example, the input sentence of "Α 之 Βis C", from the construction of the sentence, becomes (the main 忐: use ⑻, data (eight of…: the role of complement (none), data (0: } 'Again, D will go to E ”, referring to the case of the detected verb (go) Table to generate {verb (go), function ((i) (wa), data ⑼, function ⑹⑽, data (E)}. Word decomposition, noun extraction device 113, the result of morphological element analysis is opened from the attached word information The extracted word extracts the word of the noun, and sends the word to the rhetorical supplementary device 116. The input sentence judging device 114 detects the presence of "?", "Jealousy" (ka) and interrogative debris at the end of the sentence, and separates the positive sentence separately. Passive list confirmation device Η $ ❿ question to input sentence processing device 117 and send the input sentence target. The secondary 袼 list confirmation device 115 is based on the detected verbs and draws a list from a poor library, not shown, and enters The sentence formation information of the sentences matches with each other. If there is an unfilled case (inadequate), the insufficient information is sent to the questioning supplementary device 11 6. If there is no shortage, the input sentence object is transmitted to the input sentence processing device 117. The question answering device 116 will be described with reference to FIG. 8 showing its details. After analyzing the result of the input sentence accepted from the normal mode, if a defect is detected, the device confirms that the device has sent insufficient information. The short-term holding device 211 stores the short-term information in the buffer 212, and applies the Chinese National Standard (CNS) Α4 specification (2) to the paper size (2) 〇X297ij ^-

-16- 528966 五、發明説明(14 ) 將該資訊送出至輸入句處理裝置117。又,從回答模式所輸 入時’即,從單字分解―名詞抽出裝置113會傳來作為對於 不足格之詢問之回答用的單字(名詞),因此,補充裝置⑴ 就將該單字填補於不足格,且傳送成為完全之目標至輸入 句處理裝置。 輸入句處理裝置117’就如在第9圖所示,由登錄處理 裝置3U、檢索處理裝置312、4認處理裝置313、詢問處理 裝置314、動作執行處理裝置315,以及知識資料庫316,所 構成。 登錄處理裝置311,係接受於輸入句判定裝置114作為 肯定句並經由袼列表確認裝置115,反問補充裝置ιΐ6而無 不足格之完全之目標,且與已經登錄之知識相照會而登錄 新資料者。欲登錄時,即以rnew」不登錄時乃以「〇1屯 通知對答裝置118。 d 檢索處理裝置312乃於在輸入句判定裝置114視為疑問 句並且於未圖示之疑問詞檢測裝置檢測出有疑問詞之存在 時起動。主語為疑問詞時,就以{作用()、資料())檢索知識 f料庫316,而將命中(成功)之知識(目標)之主語()作為 回答。其以外時,乃以主語()探搜知識並從命中之知識 之中將相稱於疑問詞之格的内容作為回答。而該回答就傳 送至應答裝置11 8。 確認處理裝置313乃於在輸入句判定裝置114視為疑問 句並且在未圖示之疑問句檢測裝置檢測出有疑問詞之存在 時作起動。並且,將輸入句目標與知識資料庫316相照會, 本紙張尺度適用中國國家標準(CNS) A4規格(2】0X297公釐) (¾先閲讀背面之>!*事項再塡窝本頁) .裝· •訂| :線丨 -17. 528966 五、發明説明 、15 若有同一者即「yes」,而無之時就將rn〇」通知應答裝置 118 ° 詢問處理裝置3 14係解析從通常模式所接受之輸入句 的結果,檢測出不足格且經由格列表確認裝置,不足格暫 時保持裝置,而於有不足格資訊傳來之時作起動。其將以 格助詞「以」、「和」、「〜」、「(二」等所表現之不足格的構 句資Λ (作用),以事先所準備之變換規則而變換為主語、 目的地、出發地等之用詞,且送出至應答裝置118。 動作執行處理裝置315係解析輸入句之結果,檢測出指 定事先所設定之動作執行的關鍵字(動詞等)以及「請作(給 )」等之依賴的語尾時作動作。例如,輸入「請給Α先生傳 真B」時,就起動事先所設定之傳真程式而將文書b作傳送 。而後,於動作完了後乃對應答裝置118通知指定動作執行 之關鍵字。 應答裝置118乃如在第1〇圖所示,由隨聲附和作成裝置 411 ’回答作成裝置412,yes/n〇應答裝置“〗,詢問作成裝 置414’報告作成裝置415,輸出裝置416,所構成。、 隨聲附和作成裝置411若從登錄處理裝置3 ,於「一時,就將「是的,了解了」之回答句通知 於「⑽」之時,即以厂曾聽過」之回答句傳送至輸出裝 416。 、 回合作成裝置412乃將從檢索處理裝置所傳送之回 ,填外在「其為(回答)」之定型形成而作成回答句。Ye: 應答作成裝置413,於從確認處理裝置之通知為「乂^ 而置 答 /no 之 本紙張尺度適财關家標準⑽s) A4規格(2Κ)χ297^)-16- 528966 V. Description of the invention (14) The information is sent to the input sentence processing device 117. In addition, when inputting from the answer mode, that is, from word decomposition, the noun extraction device 113 will send a word (noun) as a response to an inadequate question. Therefore, the supplementary device ⑴ fills the word with the inadequacy. And send it as a complete target to the input sentence processing device. As shown in FIG. 9, the input sentence processing device 117 'is composed of a registration processing device 3U, a retrieval processing device 312, a recognition processing device 313, an inquiry processing device 314, an action execution processing device 315, and a knowledge database 316. Make up. The registration processing device 311 is a person who accepts the input sentence determination device 114 as an affirmative sentence and passes the 袼 list confirmation device 115, asks the supplementary device 6 without any incomplete target, and registers new data with the registered knowledge. . When you want to log in, that is, when you do n’t log in, the answering device 118 is notified with “〇1tun.” D The search processing device 312 detects the question sentence in the input sentence judging device 114 and detects the question word detection device (not shown). It starts when there is an interrogative word. When the subject is an interrogative word, the knowledge f database 316 is retrieved with {action (), data ()), and the subject () of the knowledge (target) that hits (success) is used as the answer. In other cases, the subject () is used to search for knowledge, and the content of the case corresponding to the interrogative word is taken as the answer from the hit knowledge. The answer is transmitted to the answering device 118. The confirmation processing device 313 is inputting Sentence determination device 114 regards the interrogative sentence and activates it when the interrogative sentence detection device (not shown) detects the existence of interrogative words. Moreover, the target of the input sentence is compared with the knowledge database 316, and the Chinese paper standard (CNS) ) A4 specification (2) 0X297 mm) (¾Read the back >! * Matters on this page first). Install · • Order |: Line 丨 -17. 528966 V. Description of the invention, 15 If there is one "Yes" If not, it will notify rn〇 ”to the responding device 118 °. The inquiry processing device 3 14 analyzes the result of the input sentence accepted from the normal mode, detects the defect and confirms the device through the cell list, and temporarily holds the device. Start when there is insufficient information. It will change the subject and purpose of the inadequate construction materials Λ (function) expressed by the case auxiliary words "to", "and", "~", "(II), etc., using the transformation rules prepared in advance. Words such as location, place of departure, etc., and send them to the response device 118. The action execution processing device 315 analyzes the result of the input sentence, and detects keywords (verbs, etc.) that specify execution of actions set in advance, and "Please do (for ) "And so on. For example, when you enter" Please fax B to Mr. A ", you start the fax program set in advance and send the document b. Then, after the action is completed, the response device 118 is transmitted. The keyword for notifying the execution of the specified action. As shown in FIG. 10, the response device 118 is composed of the accompanying creation device 411 'response creation device 412, yes / no response device', and the inquiry creation device 414 'and the report creation device 415. The output device 416 constitutes the following. The accompanying creation device 411, from the registration processing device 3, notifies the answer sentence of "Yes, I understand" to "⑽" at one time, that is, The answer sentence "Have heard it" is sent to the output device 416. The response cooperation device 412 is a response sentence sent from the search processing device, filling in the stereotype of "It is (response)" to create a response sentence. Ye: Response The creating device 413 confirms that the notification from the processing device is "乂 ^ and responds to / no of the paper standard suitable for financial and family standards⑽s) A4 specification (2Κ) χ297 ^)

----------------譬…: (^先閲讀背面之注*事項再填寫本頁) .訂— -18- 528966 A7---------------- For example ...: (^ Read the notes on the back * please fill in this page). Order — -18- 528966 A7

時,就作為「是的」而為「n〇」之時,即作為 回答句。When it is "yes" and "n〇", it is an answer sentence.

不是」之 詢問作成裝置414係將從詢問處理裳置所傳送 示不足格的主格、目的地、出發地等之用詞適用於「( )是什麼?」之定型句而作為疑問句(應答句)。報告= 置415乃將從動作執行處理裝置所通知之_字適用在、 為(關鍵字)」之定型句而作為報告(應答句)。 輸出裝置416乃將所作成之應答句輸入於未圖示之聲 音合成裝置且從未圖示之揚聲器以聲音對使用者發話而作 回答。當然,將應答句顯示在監視器等亦可。 由上述說明可知,依本專利申請之第l g樣有關之本發 明’判別使用者之自然語所產生的輸入句,而以相異種類 之對話裝置相對應,且能作因應使用者之必要而能對應更 接近於人類之對話。並且,由於設定特定之業務知識,乃 能彌補使用者之曖昧的問話,且能對使用者提示適切之資 訊。 由上述說明可知,依本申請專利之第2態樣的本發明, 可將使用者之自然語所產生的輸入句判斷其對話之狀況( 其輸入句為對於系統所發出之詢問的回答成否之發話的狀 況)’並為理解該輸入乃能依照其狀況判斷切換輸入模式而 作解析,而且,判定其輸入句之文意而執行適切之作業, 且月b將因應其執行之適切的應答句傳還於使用者而作更接 近於人類之對話。 本紙張尺度適用中國國家標準(CNS) A4規格(21〇x297公釐) -19- 、17528966 A7 五、發明説明 【圖式之簡單說明】 第1圖係表示對本發明之第1態樣有關的由電腦所產生 之自然語對話系統的一實施例整體之方塊圖。 第2圖係表示於第1圖之由電腦所產生的自然語對話系 統所使用之基本規格的一例之表。 第3圖係表示於第丨圖之由電腦所產生的自然語對話系 統所使用之業務的關鍵字•業務項目之列表的一例之表。 第4圖係表示在第丨圖之由電腦所產生的自然語對話系 統上之閒談模式對話裝置的詳細圖。 第5圖係表示在第丨圖之由電腦所產生的自然語對話系 統上之一般模式對話裝置之詳細圖。 第6圖係表示在第丨圖之由電腦所產生的自然語對話系 統上之業務模式對話裝置之詳細圖。 第7圖係表示對本發明之第2態樣有關的由電腦所產生 之自然语對話系統上的一實施例之方塊圖。 第8圖係表示在第7圖之由電腦所產生的自然語對話系 統上的反問補充裝置之詳細圖。 第9圖係表示在第7圖之由電腦所產生的自然語對話系 統上之輸入句處理裝置之詳細圖。 第1〇圖係表示在第7圖之由電腦所產生的自然語對話 系統上的應答裝置之詳細圖。 【元件標號對照】 1〇1···輸入裝置 1〇3…切換裝置 1〇2···分析裝置 刚…閒談模式對話裝置 本紙張尺度適用中國國家標準(CNS) A4規格(2】0Χ297公釐) ----------------------#…: (請先閲讀背面之:/i*事項再填寫本頁) 訂_ -20- 、18528966 A7 B7 五、發明説明 105…一般模式對話裝置 106···業務模式對話裝置 107,416···輸出裝置 108…反問裝置 109…反問裝置. 111…模式判斷裝置 112···單字分解一作用分 析裝置 113…單字分解一名詞抽 出裝置 114···輸入句判定裝置 π5···格列表確認裝置 116···反問補充裝置 117···輸入句處理裝置 Η8···應答裝置 211…不足格暫時保持裝置 212…緩衝器 213…補充裝置 311…登錄處理裝置 312…檢索處理裝置 313…確認處理裝置 314…詢問處理裝置 315…動作執行處理裝置 3 16…知識資料庫 401· 402· 403· 404· 411· 412· 413· 414· 415· 501· 502· 503· 504· 505· 506· 507· 508· 601· 602· 603· 604· 605· 規則選擇裝置 規則庫 應答生成裝置 計時器 隨聲附和作成裝置 回答作成裝置 YES/NO應答裝置 詢問作成裝置 報告作成裝置 構句資訊生成裝置 動詞之格列表資料庫 疑問句判別裝置 檢索裝置 一般知識資料庫 知識生成裝置 隨聲附加生成裝置 回答生成裝置 業務模式監視裝置 模式遷移裝置 業務特定裝置 項目監視裝置 檢索條件之收集裝置 606···格列表之順序不 處理裝置 同 請先閲讀背面之注意事項冉褀寫衣頁) -裝- 訂· :線 本紙張尺度適用中國國家標準(CNS) Α4規格(210X297公爱) 21 528966The query creation device 414 that is not "is a question sentence (response sentence) that applies the words" information (")"? . The report = setting 415 is a report (response sentence) in which the _ word notified by the action execution processing device is applied to the set sentence (keyword). The output device 416 inputs the created response sentence to a voice synthesizing device (not shown) and responds to the user with a voice from a speaker (not shown). Of course, the response sentence may be displayed on a monitor or the like. From the above description, it can be seen that according to the invention according to the first lg of the present patent application, the input sentence generated by the natural language of the user is corresponding to a different type of dialogue device, and can be used according to the needs of the user. Can correspond to dialogue closer to humans. In addition, by setting specific business knowledge, it can make up for ambiguous questions from users, and can prompt users with appropriate information. As can be seen from the above description, according to the second aspect of the present invention, the input sentence generated by the user's natural language can be used to determine the status of the conversation (whose input sentence is the answer to the query issued by the system or not). The state of the utterance) 'and to understand that the input can be switched according to its situation to determine the input mode for analysis, and determine the meaning of the input sentence and perform appropriate tasks, and month b will respond to the appropriate response sentence Return to the user and make a dialogue closer to humans. This paper size applies the Chinese National Standard (CNS) A4 specification (21 × 297 mm) -19-, 17528966 A7 V. Description of the invention [Simplified description of the drawings] Figure 1 shows the first aspect of the invention. An overall block diagram of an embodiment of a natural language dialogue system generated by a computer. Fig. 2 is a table showing an example of the basic specifications used by the computer-generated natural language dialog system shown in Fig. 1. Fig. 3 is a table showing an example of a list of keywords and business items of a business used by a computer-generated natural language dialogue system shown in Fig. 丨. FIG. 4 is a detailed diagram showing a chat mode dialogue device on the natural language dialogue system generated by the computer in FIG. FIG. 5 is a detailed diagram showing a general mode dialogue device on the natural language dialogue system generated by the computer in FIG. FIG. 6 is a detailed diagram showing a business mode dialogue device on the natural language dialogue system generated by the computer in FIG. Fig. 7 is a block diagram showing an embodiment of a natural language dialogue system generated by a computer according to a second aspect of the present invention. Fig. 8 is a detailed diagram of a rhetorical supplementary device on the computer-generated natural language dialogue system of Fig. 7. Fig. 9 is a detailed diagram showing an input sentence processing device on a natural language dialogue system produced by a computer in Fig. 7. Fig. 10 is a detailed diagram showing a response device on the natural language dialogue system produced by the computer in Fig. 7. [Comparison of component numbers] 1101 ... Input device 103 ... Switching device 102 ... Analysis device just ... Chat mode dialogue device This paper standard applies to China National Standard (CNS) A4 specification (2) 0 × 297 male %) ---------------------- #…: (Please read the following: / i * Matters before filling out this page) Order _ -20-, 18528966 A7 B7 V. Description of the invention 105 ... General mode dialogue device 106 ... Business mode dialogue device 107, 416 ... Output device 108 ... Questioning device 109 ... Questioning device 111 ... Mode judging device 112 ... Single word decomposition-action analysis device 113 ... Word decomposition and noun extraction device 114 ... Input sentence determination device π5 ... Case list confirmation device 116 ... Question supplement device 117 ... Input sentence processing device Η 8 ... Response device 211 ... Insufficient temporarily Holding device 212 ... buffer 213 ... supplementary device 311 ... register processing device 312 ... retrieval processing device 313 ... confirmation processing device 314 ... inquiry processing device 315 ... action execution processing device 3 16 ... knowledge database 401 · 402 · 403 · 404 · 411 · 412 · 413 · 414 · 415 · 501 · 502 503, 504, 505, 506, 506, 507, 508, 601, 602, 603, 604, 605, and 601. 602. 603. 604. 605. 605. Rule selection device. Rule library response generation device. Timer echo creation device response creation device. YES / NO response device. Inquiry creation device. Report creation device. Sentence formation information generating device Verb list database Interrogative sentence discrimination device Retrieval device General knowledge database Knowledge generation device Acoustic addition generation device Answer generation device Business mode monitoring device Mode migration device Business specific device Item monitoring device search device collection device 606 ··· The order of the grid list is not the same as the device. Please read the notes on the back. (Ran Yi writes the clothes page)-Binding-Binding:: The size of the thread paper is applicable to the Chinese National Standard (CNS) Α4 specification (210X297 public love) 21 528966

-*ftj/-j日听提之 A7 B7 五、發明説明(19 ) 607…檢索條件緩衝器 610…格列表抽出裝置 608…檢索/回答句生成裝置 611…業務知識資料庫 609…格列表緩衝器 本紙張尺度適用中國國家標準(CNS) A4規格(210X297公釐) (請先閲讀背面之注念事項再填寫本頁) 22-* ftj / -j A7 B7 of the day-to-day mention V. Description of the invention (19) 607 ... Search condition buffer 610 ... Grid list extraction device 608 ... Search / response sentence generation device 611 ... Business knowledge database 609 ... Grid list buffer The paper size of this device applies to the Chinese National Standard (CNS) A4 (210X297 mm) (Please read the notes on the back before filling this page) 22

Claims (1)

528966 足 A8 B8 C8 D8 、申睛專利範圍 1 ·種由電腦所產生之自然語對話系統,該電腦係至少具 有可由使用者輸入資訊之輸入裝置,而該系統係包含有 對話模式選擇裝置,係、可由性質各異之至少2種對 話模式中選擇其一者; 分析裝置,係可將輸入句作形態要素解析,並將單 子列作為資料列出者;及 刀換裝置,係可判別前述分析之結果來切換前述對 話模式者。 2·如申請專利範㈣1項之由電腦所產生之自然語對話系 、先其中5亥對話模式係至少包含由以下模式所選出之2 種模式: 閒 < 模式,该對話模式係僅依照事先所設定之規則 庫來輸出應答而不獲得資料者; 通吊'拉式,遠對話模式係判定輸入句之動詞後參照 預先準備之辭典來獲得該動狀格資訊,並從輸入句抽 出適合該格之資料,而若無適合資訊之格時,則對使用 2輪出詢問該不足格之疑問句來要求使用者進行不 ”;斗輪入以獲得使該格資訊符合項目之資料者;及 業務模式,該對話模式係參照視纟統管 需要而預先設定之項目(以下,稱為「業務用項目業二 後,從輸入句獲得符合該業務用項目之資訊,若有未填 =之項目時,乃對使用者輪出詢問該不以目之疑問句 來要求使用者進行^足資訊輸人,以獲得對應於未填補528966 Foot A8 B8 C8 D8, patent application scope 1 · A natural language dialogue system generated by a computer, the computer has at least an input device capable of inputting information by a user, and the system includes a dialogue mode selection device. 1. You can choose one of at least two dialogue modes with different properties; the analysis device can analyze the input sentence as a morphological element and list the list as the data; and the tool changer can judge the analysis As a result, the aforementioned dialogue mode is switched. 2. If the computer-generated natural language dialogue system in item 1 of the patent application, the first 5 Hai dialogue mode includes at least two modes selected from the following modes: leisure < mode, this dialogue mode is based on prior The rule base is set to output the response without obtaining the data. The “drag-and-drop”, remote dialogue mode is to determine the verb of the input sentence and refer to a dictionary prepared in advance to obtain the verbal case information. Qualified information, and if there is no suitable information format, ask the user to use "2 rounds to ask the inadequate interrogative" to ask the user to do not; Mode, this dialogue mode refers to the items that are set in advance according to the needs of the management (hereinafter referred to as "business project industry 2", and obtains information that matches the business project from the input sentence. If there are unfilled = items, It asks the user to ask questions about the non-target questions and asks the user to input enough information to obtain the corresponding unfilled information. f請先閲讀背面之注意事項再填寫本頁} 23 528966 六、申請專利範圍 之業務用項目之資料者。 3. 如申請專利範圍第2項之電腦所產生之自然語對 ’其中該業務用模式係進而區分為業務用項目^ 數業務模式。 複f Please read the precautions on the back before filling out this page} 23 528966 VI. Those who apply for business scope of patent application. 3. For example, the natural language pair generated by the computer in the scope of patent application No. 2 ′ where the business model is further divided into business items ^ business model. complex 4. 如申請專利範圍第3項之電腦所產生之自然語對話系统 ’其係進行-順序不同處理,而該順序不同處理係來昭 自使用者輸人句之分析所列出之單字資料及儲存該資 料所屬項目之貧料庫來判定前述列出之資料應收納之 業務項目。 5. 如申請專利範圍第2項之電腦戶斤產生之自然語對話系統 其係連結與前述業務相關之資料庫,以獲得該資料庫 之綱目來作為業務用項目。 6. 如申請專利範圍第2項之電腦所產生之自然語對話系統 ,其係具備一回歸功能,該回歸功能係可在前述業務模 式之對話中,即使當業務用項目尚未填補之狀態下而其 它模式啟動時,亦記憶前述未填補之業務用項目並 要詢問不足項目。 7. —種電腦所產生之自然語對話系統,該電腦係至少具有 可由使用者輸入資訊之輸入裝置,而該系統係至少包含 業務模式對話裝置,且,包含有: 獲得資訊裝置,係參照視系統管理者業務上需要而 預先設定之業務用項目來從輸入句獲得符合該業務用 項目之資訊者; 輸出疑問句裝置,係於有未填補之項目之場合,即 本紙張尺度適用中國國家標準(⑽)A4規格(21〇X297公釐) 心^664. If the natural language dialogue system generated by the computer in the scope of patent application No. 3 'is performed in a different order, the different order processing is based on the word data listed in the analysis of user input sentences and Store the poor data base of the project to which the data belongs to determine the business items that the previously listed data should be stored. 5. For example, the natural language dialogue system generated by the computer households in the scope of patent application No. 2 is to link the database related to the aforementioned business to obtain the outline of the database as a business project. 6. If the natural language dialogue system generated by the computer in the scope of patent application No. 2 is provided with a regression function, the regression function can be in the dialogue of the aforementioned business model, even when the business items have not been filled yet. When other modes are activated, the aforementioned unfilled business items are also memorized and inquiries are required. 7. A natural language dialogue system generated by a computer, the computer has at least an input device capable of inputting information by a user, and the system includes at least a business mode dialogue device, and includes: an information device, which refers to a video The system manager needs the business items that are set in advance to obtain information from the input sentence that matches the business items. The output question device is used in the case of unfilled items, that is, this paper standard applies Chinese national standards ( ⑽) A4 size (21 × 297 mm) heart ^ 66 對使用者輸出詢問該不足項目之疑問句者;及 、獲得資料裝置,係由於對使用者作不足資訊輸入之 8要求,以獲得對應於未填補之業務用項目之資料者。 .如申睛專利範圍第7項之電腦所產生之自然語對話系統 ,其中該業務用模式係進而區分為業務用項目相異之複 數業務模式。 9·如申請專利範圍第8項之電腦所產生之自然語對話系統 其係進行一順序不同處理,而該順序不同處理係參照 自使用者輸入句之分析所列出之單字資料及儲存該資 料所屬項目之資料庫來判定前述列出之資料應收納之 業務項目。 1〇·如申請專利範11第7項之電腦所產生之自然語對話系統 其係連結與前述業務相關之資料庫,以獲得該資料庫 之綱目來作為業務用項目。 U.如申請專利範圍第7項之電腦所產生之自然語對話系統 ,其係具備一回歸功能,該回歸功能係可在前述業務模 式之對話中,即使當業務用項目尚未填補之狀態下而其 它模式啟動時,亦記憶前述未填補之業務用項目並視需 要詢問不足項目。 12 ·種電腦所產生之自然語對話系統,該電腦係至少具有 可由使用者輸入資訊之輸入裝置,而,該系統係至少包 含有:業務誘導裝置,係可於滿足已經過預先設定之時 間與已對話預先設定之次數中至少一條件的期間未進 行業務用對話時,進行預先所設定之詢問者。 本紙張尺度翻中關家標準(CNS) M規格⑵QX297公楚) (請先閱讀背面之注意事項再填寫本頁)The user outputs a question asking the user about the insufficient item; and The device for obtaining data is a person who obtains the data corresponding to the unfilled business item due to the 8 requirements for the user to enter insufficient information. . For example, the natural language dialogue system produced by the computer in the seventh item of the patent scope of Shenyan, where the business model is further divided into multiple business models with different business items. 9. If the natural language dialogue system generated by the computer of the patent application No. 8 is processed in a different order, the different order processing refers to the word data listed in the analysis of the user input sentence and stores the data The database of the project to determine the business items that should contain the data listed above. 10. The natural language dialogue system generated by the computer as claimed in item 7 of Patent Application No. 11 is to link a database related to the aforementioned business to obtain the outline of the database as a business project. U. If the natural language dialogue system generated by the computer of the 7th scope of the patent application has a regression function, the regression function can be in the dialogue of the aforementioned business model, even when the business project has not been filled yet. When other modes are started, the aforementioned unfilled business items are also memorized and inquiries are made as needed. 12 · A natural language dialogue system generated by a computer, the computer having at least an input device capable of inputting information by a user, and the system including at least: a business induction device, which can meet a predetermined time and If at least one of the pre-set times of the dialog has been talked with, the business inquiry is not performed, and the inquiry set in advance is performed. The size of this paper is translated into Chinese Standards (CNS) M specifications (QX297) (please read the precautions on the back before filling this page) 25 13.種電腦與使用者之自然語對話方法,該電腦係至少具 有可由使用者輸入資訊之輸入裝置,而該方法係包含有 對話模式選擇步驟,係可由性質各異之至少2種對 話模式中選擇其一者; 分析步驟,係可將輸入句作形態要素解析,並將單 字列作為資料列出者;及 切換步驟,係可判別前述分析之結果來切換前述對 活模式者。 14·如申請專利範圍第13項之電腦與使用者之自然語對話 方法,其中該對話模式係至少包含由以下模式所選 2種模式: 閒談模式,該對話模式係僅依照事先所設定之規則 庫來輸出應答而不獲得資料;25 13. A natural language dialogue method between a computer and a user, the computer having at least an input device capable of inputting information by the user, and the method including a dialogue mode selection step, which can be performed by at least two dialogue modes with different properties Choose one of them; the analysis step is to analyze the input sentence as a morphological element and list the single word as the data; and the switching step is to determine the result of the analysis to switch the active mode. 14. The natural language dialogue method between the computer and the user according to item 13 of the patent application scope, wherein the dialogue mode includes at least two modes selected by the following modes: Chat mode, the dialogue mode is only in accordance with the rules set in advance Library to output responses without obtaining information; 通常模式,該對話模式係判定輸入句之動詞後參照 預先準備之辭典來獲得該動狀格資訊,並從輸入句抽 出適合該格之資料,而若無適合資訊之格時,則對使用 者輸出詢問該不足袼之疑問句來要求使用者進行不足 資料輸入,以獲得使該格資訊符合項目之資料;及 業務模式,該對話模式係參照視系統管理者之業務 需要而預先設定之項目(以了,稱為「業務用項目」) 後,從輸入句獲得符合該業務用項目之資訊,若有未填 補之項目時,乃對使用者輸出詢問該不足項目之疑問句 來要求使用者進行不足資訊輸人,賴得對應於未填補 本紙張尺度適用中國國家標準(CNS) A4規格(210X297公爱) 26 _清專利範圍 之業務用項目之資料。 15·如申請專利範圍第14項t電腦與使用者之自然語對話 、、-中該業務用模式係、進而區分為業務用項目相異 之複數業務模式。 、 16·如申凊專利範圍第15項之電腦與使用者之自然語對話 方法’其更包含有順序不同處理步驟,係參照自使用者 輪入:之分析所列出之單字資料及儲存該資料所屬項 =之貧料庫來判定前述列出之f料應收納之業務項目 17·如申請專利範圍第14項之電腦與使用者之自然語對話 方法其更包含有獲得綱目步驟,係連結與前述業務相 I之貝料庫’以獲得該資料庫之綱目來作為業務用項目 18·如申請專圍第14項之錢與使用者之自然;語對話 方法’其更包含有回歸步驟,係可在前述業務模式之對 話中’即使當業務用項目尚未填補之狀態下而其他模式 啟動時’亦記憶前述未填補之業務用項目並視需要詢問 不足項目者。 19· -種電腦與使用者之自然語對話方法,該電腦係至少具 有可由使用者輸入資訊之輸入裝置,而該方法係至少包 含有業務模式對話步驟,且,包含有: 獲得資訊步驟,係參照視系統管理者業務上需要而 預先設定之業務用項目來從輸入句獲得符合該業務用 項目之資訊者; 528966 、申請專利範圍 輸出疑問句步驟,係於有未填補之項目時,即對使 用者輸出詢問該不足項目之疑問句者;及 θ獲得資料步驟,係要求使用者輸入不足資訊,以獲 仵對應於未填補之業務用項目之資料者。 2 0 ·如申請專利範圍第19項之電腦與使用者之自’然語對話 方法,其中該業務用模式係進而區分為業務用項目相里 m 之複數業務模式。 〃 Λ 21.如申請專利範圍第19項之電腦與使用者之自然語對話 方法’其更包含有順序不同處理,係參照自使用者輪 :之分析所列出之單字資料及儲存該資料所屬項目 貧料庫來判定前述列出之資料應收納之業務項目者。 22·如申請專利範圍第19項之電腦與使用者之自然語對為 方法’其更包含有獲得綱目步驟’係連結與前述業務相 資料庫’以獲得該資料庫之綱目來作為業務用項目 η如申請專·圍第19項之錢與❹者之自然語對話 其更包含有回歸步驟’係可在前述業務模式之對 即使當業㈣項目尚未填補之㈣下而其他模式 2時’亦記憶前述未填補之業務用項目並視需要詢問 不足項目者。 種電腦與使用者之自然語對話方法,該電腦係至少具 可由使用者輸人資訊之輸人裝置,而該方法係 §一業務誘導步驟,係可於滿足已經過預先設定之二 與已對話預先設定之次數中至少—條件的期間未進Γ '、、申清專利範圍 業務用對話時,進行預先所設定之詢問者。 25· 一種電腦可讀取之記錄媒體,其係記錄有可執行申請專 3 圍第13項至第24項之f腦與使用者之自然語:話 方法之程式者。 26·-種電腦所產生之自然語對話系統,其係包含有: 者·模式㈣係·對話之狀況來切換輸入模式 單子刀解裝置,係對輸入句作形態要素分析者; 作用刀析裝置,係將前述形態要素分析之結果作構 句刀析且生成附有表示構句上之格的標籤之單字列表( 以下’稱為「構句表現資料」)者; 反門補充裝置’係輸入對於該系統所輸出之疑問句 的回答句而即判斷時,從前述形態要素分析之結果列出 品詞為名詞之單字且抽出作為回答者; 輸入句判定裝置,係可於判定輸入句後,從將前述 構句表現資料作為知識而儲存在資料庫之登錄處理、檢 索已儲存在前述資料庫之知識之檢索處理、催促使用者 輸入構句不元全之輪入句不足格之反問處理以及使預 先没疋之功能啟動之功能執行處理等選擇至少一輸入 句處理者; 輸入句處理裝置,係依照前述輸入句判定而進行輸 入句處理者;及 應答裝置’係視前述輸入句處理裝置之結果而生成 輸出句並輸出者。 本紙張尺度適用中國國家標準(CNS) M規格(21〇>(297公楚) 29 528966 申請專利範園 27. 如申請專利範圍第26項之由電腦所產生之自然語對$ 系統,其中該輸入模式係至少由通常模式及輸入對於兮 系統所輸出之疑問句的回答句時的回答模式之2種所構 成。 28. 如申請專利範圍第26項之由電腦所產生之自然語對話 系統,其中該輸入句係至少由以鍵盤輸入之正文句及聲 音認識裝置所正文化之文章所選出之其一。 29·如申請專利範圍第26項之由電腦所產生之自然語對話 系統,其中該對話之狀況,係以在剛才該系統是否對該 對話之對象輸出疑問句而作判斷。 30· —種電腦與使用者之自然語對話方法,該電腦係至少具 有可由使用者輸入資訊之輸入裝置,而該方法係包含有 切換步驟,係判斷對話之狀況來切換接受輸入之模 式者; 料( 理解步驟,係依前述模式解析輸入句後,再構成為 由附有解析結果標籤之單字所構成之列表構造之資 以下%「知識化資料」)並理解該輸入句者; 之 判斷步驟,係分析前述知識化資料且判斷由知識〜 傳杈知硪之確認、知識之檢索以及預先設定之功能動 作的依賴而所選出之; 執仃步驟’係執行對應前述判斷結果之作業者;及 回t步驟,係生成對應前述作業執行之回答句後, 輪出並回答者。 本紙張尺度適用中 國國 格(210X29^jy 30 ^>28966 佟. 無明欠。、 %提 正之 、申請專利範圍 2Γ> .如申請專利範圍第30項之電腦與使用者之自然語對話 方法’其中該輸入模式係至少由通常模式及對於該系統 所輪出之疑問句有輸入回答句時的回答模式之2種模式 所構成。 32.如申請專利範圍第3()項之電腦與使用者之自然語對红 方法’其中該輸入句係至少為經由自鍵盤所輸入之正文 句及聲音認識裝置所正文化之文章所選出之其一。 33·如_請專利範圍第3G項之電腦與使用者之自然:語 方法,其中該對話之狀況係以在剛才該系統對於該對話 之對象是否輸出疑問句而作判斷。 34·如申請專利範圍第30項之電腦與使用者之自然語對話 方法,其中該標記係參照至少由預先作成之辭典,以及 將輸入句作構句解析之結果所選出之其一而進行。 乃種電腦可讀取之記錄媒體,其係記錄有執行^請專利 範圍第30項至第34項之電腦與使用者之自然語對話 法之程式。 訂 方 本紙張尺度適用中國國家標準(CNS) A4規格(210X297公爱) 31Normal mode, the dialogue mode is to determine the verb of the input sentence and to refer to a dictionary prepared in advance to obtain the verbal case information, and to extract the data suitable for the case from the input sentence. Output a question asking the inadequacy to ask the user to enter inadequate data to obtain data that matches the grid information with the project; and a business model, which is a dialog mode that refers to items set in advance based on the business needs of the system manager (using (Referred to as "business items"), and then obtain information that matches the business item from the input sentence. If there are unfilled items, a question sentence asking the user for the insufficient item is output to the user to ask the user for insufficient information. The loser corresponds to the information on the business items that are not covered by the paper standards that apply the Chinese National Standard (CNS) A4 specifications (210X297 public love). 15. If the natural language dialogue between the computer and the user in item 14 of the scope of patent application, the business model is divided into multiple business models with different business items. 16. The natural language dialogue method between the computer and the user, such as the 15th in the scope of the patent application, which includes different processing steps in order, referring to the analysis of the word data listed in the user's turn: and storing the The data item belongs to the poor material library to determine the business items that should be included in the f materials listed above.17. If the natural language dialogue method between the computer and the user in the application for the scope of patent application No. 14 includes the steps of obtaining the outline, it is a link The shellfish database with the aforementioned business I 'to obtain the outline of the database as a business item 18. If you apply for the money of the 14th item and the nature of the user; the dialogue method' It also includes a regression step, In the dialogue of the aforementioned business model, 'even when the business model has not yet been filled and other models are activated', the aforementioned unfilled business project is memorized and the inadequate items are asked if necessary. 19. · A natural language dialogue method between a computer and a user. The computer has at least an input device capable of inputting information by the user. The method includes at least a business mode dialogue step, and includes: an information obtaining step, Refer to the business items that are set in advance according to the business needs of the system manager to obtain information that matches the business items from the input sentence; 528966, the steps of applying for a patent scope and outputting interrogative sentences are used when there are unfilled items. The person who outputs the question asking the inadequate item; and the step of obtaining data, which requires the user to input insufficient information to obtain the data corresponding to the unfilled business item. 2 0. If the method of natural language dialogue between computer and user in item 19 of the scope of patent application, the business mode is further divided into a plurality of business modes in the business project phase m. 〃 Λ 21. If the method of natural language dialogue between computer and user in item 19 of the scope of patent application, it also includes different order processing, referring to the analysis of the word data listed in the user wheel: and storing the data. Project poor materials library to determine the business items that should be included in the information listed above. 22 · If the natural language pair between the computer and the user in item 19 of the patent application is the method 'It further includes the step of obtaining the outline', which is to link with the aforementioned business-related database 'to obtain the outline of the database as a business item η If you apply for the money of item 19 and talk with the natural language of the person, it also includes a regression step 'can be in the above business model pair even if the business project has not yet been filled and other models 2' also Remember the aforesaid unfilled business items and ask those who are inadequate if necessary. A natural-language dialogue method between a computer and a user. The computer is at least an input device capable of inputting information by the user. The method is a business induction step. At least one of the preset number of times-the conditional period has not been entered, and when the patent-scope business dialogue has been declared, a preset inquiry is performed. 25 · A computer-readable recording medium, which records the executable language of the 13th item to the 24th item of the brain and the user's natural language: words method program. 26 ·-A natural language dialogue system produced by a computer, which includes: the state of the mode, the system of the system, and the state of the dialogue to switch the input mode; a unit for analyzing the input sentence; The result of the analysis of the aforementioned morphological elements is a sentence analysis and a list of words (hereinafter referred to as "sentence formation performance data") with a label indicating a case on the sentence is formed; an anti-door supplementary device is an input for When the answer sentence of the question sentence output by the system is judged, from the result of the analysis of the above-mentioned morphological elements, the word with the word as a noun is listed and extracted as the answerer; the input sentence judging device can judge the input sentence from the previous sentence. Sentence formation performance data is stored in the database as knowledge, registration processing, retrieval processing to retrieve the knowledge that has been stored in the aforementioned database, urged users to enter inadequate round-in sentence constructions, and question processing to make incomplete启动 The function to start the function execution processing, etc. select at least one input sentence processor; the input sentence processing device is based on the input sentence judgment An input sentence processor; and a response device ' is an output sentence generated and outputted by a result of the input sentence processing device. This paper size applies the Chinese National Standard (CNS) M specification (21〇 > (297 Gongchu) 29 528966 Patent Application Park 27. For example, the natural language pair system generated by a computer such as the 26th scope of the patent application, where The input mode is composed of at least two modes: a normal mode and a response mode for inputting a response sentence to a question sentence output by the system. 28. For example, a natural language dialogue system generated by a computer according to item 26 of the scope of patent application, The input sentence is at least one selected by a text sentence inputted by a keyboard and an article of the culture recognized by the sound recognition device. 29. For example, a natural language dialogue system generated by a computer, such as the scope of application for patent No. 26, where: The status of the dialogue is judged by whether the system has just outputted question sentences to the object of the dialogue. 30 · —A natural language dialogue method between a computer and a user, the computer has at least an input device capable of inputting information by the user, The method includes a switching step, which is to judge the status of the dialogue to switch the mode of accepting input; After the input sentence is parsed by the pattern, it is reconstituted as a list structure composed of words with analysis result tags ("information data") and understand the input sentence; the judgment step is to analyze the aforementioned knowledge data and Judgment is selected by the knowledge-to-knowledge confirmation, knowledge retrieval, and the dependence of pre-set functions and actions; the execution step 'is to perform the operator corresponding to the aforementioned judgment result; and the step t is to generate the corresponding correspondence to the aforementioned After the answer sentence for the execution of the job, turn out and answer. The paper size applies to the national standard of China (210X29 ^ jy 30 ^ > 28966 无. No owed.,% Correction, patent application scope 2Γ >. A 30-item natural language dialogue method between a computer and a user, wherein the input mode is composed of at least two modes: a normal mode and a response mode when there is an input answer sentence for a question sentence rotated by the system. Method of natural language redemption of computer and user in item 3 () of the patent scope 'where the input sentence is at least input from a keyboard One of the articles selected by the text and sound recognition device. 33. If _ please ask the computer and the user's natural: language method in item 3G of the patent, where the status of the dialogue is based on the system just now. The subject of the dialogue judges whether or not the question is output. 34. For example, the natural language dialogue method between the computer and the user in the scope of patent application No. 30, wherein the mark refers to a dictionary prepared at least in advance and the input sentence is analyzed as a sentence structure. The result is one of the selected ones. It is a computer-readable recording medium that records a program that implements the natural language dialogue method between a computer and a user in the scope of the patent claims 30 to 34. Order This paper size applies to China National Standard (CNS) A4 (210X297 public love) 31
TW089119245A 1999-07-06 2000-09-19 Natural language dialog system by computer, method and recording medium TW528966B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP11191311A JP2001022763A (en) 1999-07-06 1999-07-06 Natural word conversation by computer
JP29911899A JP2001117923A (en) 1999-10-21 1999-10-21 Natural word interactive system for computer

Publications (1)

Publication Number Publication Date
TW528966B true TW528966B (en) 2003-04-21

Family

ID=26506620

Family Applications (1)

Application Number Title Priority Date Filing Date
TW089119245A TW528966B (en) 1999-07-06 2000-09-19 Natural language dialog system by computer, method and recording medium

Country Status (2)

Country Link
TW (1) TW528966B (en)
WO (1) WO2001003006A1 (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110851579B (en) * 2019-11-06 2023-03-10 杨鑫蛟 User intention identification method, system, mobile terminal and storage medium

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6217871A (en) * 1985-07-17 1987-01-26 Agency Of Ind Science & Technol Topic control system in computer system
JPS63106042A (en) * 1986-10-23 1988-05-11 Nippon Telegr & Teleph Corp <Ntt> Method for recognizing intention of conversation text
JP2596260B2 (en) * 1990-07-18 1997-04-02 三菱電機株式会社 Correspondence table use database search device
JPH04186435A (en) * 1990-11-21 1992-07-03 Agency Of Ind Science & Technol Interaction control method and interaction control device
JPH04332084A (en) * 1991-05-07 1992-11-19 Nippon Telegr & Teleph Corp <Ntt> Automatic information offering method
JP3454897B2 (en) * 1994-01-31 2003-10-06 株式会社日立製作所 Spoken dialogue system
JPH08147310A (en) * 1994-11-25 1996-06-07 Nippon Telegr & Teleph Corp <Ntt> Request prediction type information providing service device
JP2809136B2 (en) * 1995-06-26 1998-10-08 日本電気株式会社 Natural language dialogue device
JPH0991309A (en) * 1995-09-28 1997-04-04 Toshiba Corp Multiple-data-base retrieval device
EP0829811A1 (en) * 1996-09-11 1998-03-18 Nippon Telegraph And Telephone Corporation Method and system for information retrieval
JPH1097533A (en) * 1996-09-24 1998-04-14 Mitsubishi Electric Corp Language processor
JPH11203325A (en) * 1998-01-16 1999-07-30 Tadamitsu Ryu Database generating method, record medium in which its program is stored and record medium recording group of data generated by the generating method
JP3178426B2 (en) * 1998-07-29 2001-06-18 日本電気株式会社 Natural language dialogue system and natural language dialogue program recording medium

Also Published As

Publication number Publication date
WO2001003006A1 (en) 2001-01-11

Similar Documents

Publication Publication Date Title
KR101883185B1 (en) Robot automatic counselling method and system with determined scenario using machine-learning
CN107209759B (en) Annotation support device and recording medium
US8548805B2 (en) System and method of semi-supervised learning for spoken language understanding using semantic role labeling
CA2520420C (en) Apparatus and method for spoken language understanding by using semantic role labeling
US8170880B2 (en) Systems and methods for monitoring speech data labelers
AbuShawar et al. Usefulness, localizability, humanness, and language-benefit: additional evaluation criteria for natural language dialogue systems
CN108829682B (en) Computer readable storage medium, intelligent question answering method and intelligent question answering device
US20080071533A1 (en) Automatic generation of statistical language models for interactive voice response applications
CN111651571A (en) Man-machine cooperation based session realization method, device, equipment and storage medium
McTear et al. Handling errors and determining confirmation strategies—an object-based approach
JPH02270067A (en) Inteligent inquiry system
CN114357126A (en) Intelligent question-answering system
US11610113B2 (en) System and method for understanding questions of users of a data management system
KR101333485B1 (en) Method for constructing named entities using online encyclopedia and apparatus for performing the same
US11314534B2 (en) System and method for interactively guiding users through a procedure
TW528966B (en) Natural language dialog system by computer, method and recording medium
JP7216627B2 (en) INPUT SUPPORT METHOD, INPUT SUPPORT SYSTEM, AND PROGRAM
CN113870478A (en) Rapid number-taking method and device, electronic equipment and storage medium
JP3910823B2 (en) Questionnaire analysis apparatus, questionnaire analysis method and program
JP2001117923A (en) Natural word interactive system for computer
Weerabahu et al. Ontology-Based Question Answering System for Banking Domain
Wang et al. A conversational agent for food-ordering dialog based on VenusDictate
Arnab et al. Shohojogi: an automated Bengali voice chat system for the banking customer services
CN116012098A (en) Order generation method and device based on network chat information
CN117743557A (en) Intelligent customer service method, intelligent customer service device, server and storage medium