201234290 六、發明說明： 【發明所屬之技術領域】 本發明係關於商業情報領域，尤其係關於獲得平面資 料的階層資訊之方法及設備。 【先前技術】 近年來’商業情報(Business Intelligence，BI)技術提 供企業全面性的商業資料相關服務，例如執行資料分析、 實施資料採集、產生資料報告、顯露資料法則等等。利用 =析資料並且獲得報告，幫助企業更有效率做出商業決 策。在商業情報技術當中，資料的維度化與階層化為後續 運用立體模型進行資料分析的基礎。 圓i顯不夕維度與多階層資料的立體模型的範例。在 此範例中，關於產品銷售的資料延著三個軸，也就是時間 (X軸）、地點(y軸）以及產品(2軸），組織成立體圖，用於描 述銷售量與時間、地點和產品之間的函數關係。進一步， 銷售量資料的結構延著每一維度分成複數個階層，如此可 根，^層執行資料分析與管理。例如在地關維度上，銷 售里資料的結構分成每一塊大陸上的銷售量；針對每一塊 ^陸，進一步區分成每個國家的銷售量；針對每個國家， 隨需求按照省份、城市料來區分。_地，針對時 度，其結構依照需求分成年、季、月、日料：針對產。 進—步根據產品的類別、系列、型號等等來區二 &康廷些維度化與階層化的資料，使用立體模 ^進行⑽P (On Line Analysis p繼ss，線上分析處貝= 刀斤與操作’如此根據使用者需求，呈現來自每—維度與 201234290 階層的整合資訊。 從上述範例可了解，資料的維度化和階層化對於商業 情報當令的資料模型化與分析提供顯著的方便性。除了一 般階層化的企業資料以外，吾人進一步想要對於其他資料 套用商業情報中的分析與操作方法。不過在許多領域當 中，例如在臨床領域中，資料仍舊以「平面」方式組織二 儲存圖2顯不現有臨床資料的範例。在圖2的範例中， 電子醫療δ己錄這類臨床資料的典型範例包括許多種資 料，例如病患的主要癥狀、診斷結論、治療等等。吾人可 了解，所有這些資料都利用臨床術語以平面方式密集排列 記，，並未顯示資料與資料的階層資訊間之關係，而、這剛 好是情報分析當巾立體㈣化與〇LAp操作的基礎。許多 類f的平面資料也存在於其他商業資料中，由於缺乏階層 f j ’運用現有的情報模型化與操作方法難以對這種 Γ行步分析與管理’如此限制了資料的系統化與情報 直赂2 ’吾人想要在現有平面資料上執行處理，以獲得 d層貝訊’幫助在平面f料上進行後續分析與管理。 【發明内容】 層資i發明係針社述問題所提出’以獲得平面資料的階 階層資訊之方t &樣’其提供—種獲得平面資料的 集的至次祖5將來自該平面資料内—相同資料 i一樹項目’映射至由—結構化術語系統所形成 'α 的至少一節點；獲得該上述樹狀結構内的 201234290 至少一子樹狀結構，該至少一子樹狀結構的每一者採用該 至少一節點做為其所有葉片節點；從該至少一子樹狀結構 當中選擇一目標樹狀結構；以及獲得該目標樹狀結構内的 階層資訊。 根據本發明的第二態樣，其提供一種獲得平面資料的 階層資訊之設備，包括··一節點映射單元，其設置成將來 自該平面資料内一相同資料集的至少一資料項目，映射至 由一結構化術語系統所形成之一樹狀結構内之至少一節 點；一子結構獲得單元，其設置成獲得該上述樹狀結構内 之至少一子樹狀結構，該至少一子樹狀結構的每一者採用 該至少一節點做為其所有葉片節點；一目標結構選擇單 元，其設置成從該至少一子樹狀結構當中選擇一目標樹狀 結構；以及一階層資訊獲得單元’其設置成獲得該目標樹 狀結構内的階層資訊。 運用本發明的方法及設備，可從平面組織資料當中獲 得資料項目之間的階層資訊，如此幫助後續在平面資料上 執行分析與管理。 【實施方式】 本說明書將結和詳細範例來說明本發明的具體實施 例。吾人應了解，所說明的範例僅為例示，不應視為本發 明範疇的限制。 如上述’本發明提供這種方法及設備，以獲得平面資 料的階層資訊。不過這種資料本身只包含複數個以平面方 201234290 式組織的資料項目，並不知道每一資料項目之間的關係， 其中該等複數個資料項目通常以術語的形式記錄在該資 料所屬襴位内。因此，要獲得平面資料的階層資訊需要有 外部結構化術語系統的幫助，這種結構化術語系統應該在 該資料所屬欄位内記錄規範詞彙，並且用階層形式組織這 些詞彙，以指出許多詞彙之間的類別與從屬關係。 下列將採用臨床領域中的臨床資料與結構化術語系 統當成範例，來說明本發明的具體實施例。 針對臨床術語系統的選擇，系統化醫學術語系統 (Systematized Terminology system of Medicine » SNOMED) 是目前廣泛使用的術語系統，其提供涵蓋大多數臨床資訊 領域方面醫療術語的系統組織電腦可處理集合，例如疾 病、發現、手術、微生物、藥品等。這樣能夠用一致的方 式索=、儲存、取得以及集十遍及專業及醫療保健單位的 臨床> 料。同時幫助組織醫療記錄的内容，減少資料擷 取、編碼方式以及用於病患臨床照護與用於研究之間的差 異。 e 其疋，SN〇MED涵蓋超過365,000個臨床詞彙，並 都指定有唯一「的數字碼、唯-的名稱(就i「完 19個上以及一個說明」。上述複數個詞彙組織成 藥物的構，包含關於臨床程序的辭彙階層、關於 階階層都具有許多分類子階層，例如：藥i 曰，kBa床失調相關詞彙可根據身體部位、病因等等 201234290 來分類，如此獲得進-步分類的階層。—個階層之内或跨 階層的不同詞彙使用大約1，偏娜個「關係」連结在一 起。如此SNOMED減朗_形成—娜構化^語系 統。在此術語系統當中，若只有考慮詞囊之間的「從屬」 關係，則可獲得具有樹狀結構的術語關係圖，其中每一術 語都是該樹狀結構的節點；並且圖中節點之間^連接線代 表郎點之間的從屬關係。在不失去一般性之下，可假設存 在一種最常見的概念，用來當成所有詞彙的根節點。通 此跟節點蚊為「Thing」，如此所有節點義接至根節點 Thing」成為其子卽點(child nodes)。如上述，因為從不 同觀點來看可在詞彙之間執行分類，每一節點都有多個子 節點以及多個親代節點(parent n〇(jes)。 根據上述SN0MED的特徵，採用SN〇MED當成結構 化術語系統來描述臨床詞彙間之階層關係是較佳選擇。不 過吾人了解’臨床術語系統的選擇並不限於SN〇meD， 也可使用已經開發或未來將開發的正常化與結構化術語 系統’例如MedDRA術語系統。這種術語系統可從不同 觀點與不同態樣形成樹狀結構，如此表達出代表詞彙的節 點間之關係。 至於在其他領域的資料，例如生物物種的資料、化學 領域内的資料等’也存在對應的結構化術語系統。如上 述’這些結構化術語系統可將領域内的標準詞彙組織成為 樹狀結構形式。 為了詳細說明’底下結合代表性臨床資料以及 SNOMED術語系統’來說明本發明的具體實施例。 201234290 圖3顯示根據本發明具體實施例的方法之流程圖。如 圖3内所示，根據本發明具體實施例一種獲得平面資料的 階層資訊之方法包括：步驟31，其中將來自該平面資料内 相同資料集的至少一資料項目，映射至由一結構化術語系 統所形成之一樹狀結構内的至少一節點；步驟32，獲得該 樹狀結構内至少一子樹狀結構，該至少一子樹狀結構的每 一者採用該至少一節點做為其所有葉片節點；步驟33，從 該至少一子樹狀結構當中選擇一目標樹狀結構；以及步驟 34，獲得該目標樹狀結構内的階層資訊。 尤其是在步驟31内，平面資料内的資料項目定位至 該結構化術語系統所形成的樹狀結構。如此首先從該平面 資料中擷取資料集，如此獲得該資料集内複數個資料項 目，如此分析來自相同資料集的資料項目，並且反映出來 自相同維度的資訊。例如在採用電子醫療記錄當成範例的 圖2所示之臨床資料内，每一垂直欄都視為來自一個維度 的資料集反映醫療記錄資訊。尤其是，第二垂直欄的資料 集内每一資料項目都用於說明病例的主要徵狀，第七垂直 欄的資料集内每一資料項目用來說明病例的診斷結論，以 及第八垂直攔的資料集内每一資料項目用於說明病例的 治療情況。因此，要採用來自相同垂直欄(也就是相同資 料集）的資料項目當成後續步驟分析的目標。 接下來，為了獲得複數個資料項目，將每一資料項目 映射至該結構化術語系統内的辭彙。在一個具體實施例 内，該平面資料為臨床資料，並且該結構化術語系統為上 述SNOMED術語系統。目前許多臨床資料已經採用 201234290 SNOMED術語系統内的標準術語記錄臨床資訊，並且某 些甚至直接採用SNOMED術語系統内詞彙的代碼，來記 錄與儲存資料。在此情況下，利用僅執行詞彙或代碼的搜 哥與配對，就可實現將臨床資料内的資料項目映射至 SNOMED術語系統内的辭彙。在臨床資料並非用規範詞 彙記錄的情況下，可另外執行資料項目與詞彙之間的字串 配對與模糊邏輯配對。在某些具體實施例内，有時也參考 術語系統内詞彙的解釋或說明當成協助。對於其他内容的 平面資料而言，同樣平面資料已經用結構化術語系統内詞 彙或代碼記錄的情況下，利用在詞彙或代碼上執行搜尋與 配對，直接實現將資料項目映射至詞彙；在平面資料並未 記錄於規範詞彙的情況下，可另外執行模糊邏輯配對。此 外，還有許多方法可用於業界内詞彙匹配，並且精通技術 人士可以此基礎選擇適當的方法，來執行資料項目與詞彙 的配對與映射。如此，將每一獲得的資料項目映射至結構 化術語系統内的辭彙。 另外因為如上述，因為結構化術語系統根據階層組織 祠彙，藉此形成術語的樹狀結構，因此採用對應至個別資 料項目的辭彙當成樹狀結構内的節點。因此資料項目已經 定位至該樹狀結構内。 士圖4A顯示根據本發明具體實施例的樹狀結構圖。該 構例示顯示SN〇MED術語系統所形成樹狀結構的 一部=，其中每一節點對應至一個詞彙，含箭頭的連接線 顯示節點之間的親子關係’並且「Thing」為整個樹狀結 構的根節點。透過上述步驟31内的映射，資料項目定位 至樹狀結構内的特定節點。在圖4A内，從資料項目映射 201234290 的節點顯示為節點八、6、〇0、丘和1?。 接下來’本具體實施例的方法前往步驟32，在上述樹 狀結構内發現，採用從資料項目映射的節點當成其全部葉 片節點的至少一子樹狀結構。仍舊參閱圖4A，步驟32的 處理為找出整個樹狀結構内的至少一子樹狀結構，每一該 子樹狀結構都採用節點A-F當成其全部葉片節點。 人 若要決定候選的子樹狀結構’需要運用到樹狀結構内 節點之間的連接關係。 在一個具體實施例内，形成該樹狀結構的結構化術語 系統（例如SNOMED)以連結開放式資料（linked 〇pen data’LOD)形式推出。在此形式中，該樹狀結構内節點之 間的關係全都用RDF triples的格式來描述與儲存。如同精 通技術人士所熟知’ RDF triple運用 <主詞，述詞，受詞> 的形式來表達許多涵義與關係。節點A和B的從屬關係(或 稱為親子關係)可用RDF triple表示為<n〇deA，subClassOf， nodeB>。針對語義型語言，在l〇D資料内具有owl:Thing 的概念；該LOD資料内的每一個別項目都是其成員，或 稱為其子節點。因此，若要查詢LOD内節點「childNode」 的親代節點，可運用下列SPARQL查詢： Select ?parentNode where {?parentNode rdfs:subClassOf <childNode>}，如此獲得該親代節點之值。也可用類似方 式查詢已知親代節點的子節點。在此情況下，透過核心述 詞subClassOf就可簡單獲得節點之間的親子關係。在其他 具體實施例内，該結構化術語系統用其他指定格式儲存。 因此’利用擷取以其他指定格式撰寫的從屬關係上之說 201234290 明 ，可獲得該舰結構㈣點之_親子關係 Μ內上可獲得節點之間的親子關係、可在該樹狀社 :=r上或往下移動，如此可透過這種移m 執行i下：趙二狀:構的根節點Thi"g 應節點，並且這種路 二—行往上移:施::處:期I:二往2 過上述SMRQL查詢獲得具有「—C二0f」 該葉片節點的至少-親代節點;然後 二=ΡΪ ·依序獲得具有較高階的上代節點， 點的單一路徑= 二:該葉片節點到該根節 徑之間的共用節點，如此:併;等=徑找二= 從該葉片節點到該根節點Thing的子樹藉此獲得 心圖4β顯示根據本發明的具體實施例，從圖4八内該 的羊^所獲得的子樹狀結構。如圖4Β内所示，該獲^ 、樹狀結構屬於圖4Α的該樹狀結構一部分；該 結構的:?有葉片節點都是從資料項目映射的節點A_F，並 點！就是™ng。不過’從觀察由該根節點 二=片節點A-F的此子樹狀結構中可發現，該子樹狀 二不是唯一採用節點Α_ρ當成葉片節點的子結構，也包 含另一個子樹狀結構’例如使用節點U當成根節點的另 12 201234290 一子樹狀結構。也就是，利用在葉片節點A-F與該終極根 節點Thing之間移動，可決定多個子樹狀結構，並且從特 定觀點或態樣中，這些子樹狀結構仍舊可反映出節點 之間潛在階層關係。因此，該獲得的多子樹狀結構之一可 依需求決定當成該目標樹狀結構，並且由該目標樹狀結構 反映出節點的階層資訊，如步驟33所示。 在一個具體實施例内，為了讓最終獲得的階層資訊更 有關聯’需要進一步篩選該等獲得的多子樹狀結構，並從 其中選擇相對「小型」的樹狀結構來反映出階層資訊，因 為在具相對「小型」結構的階層樹内，節點之間具有較緊 密的關聯性’更能夠反映出該樹狀的特定分類與主題。 底下將結合圖4B内顯示的子樹狀結構範例，來說 上述選擇處理。 在一個具體實施例内，運用兩個步驟來從多個 並選擇。首先，針對圖4B内所示從葉片節點 姓構）Zlhmg的子齡結構(在_為$—子樹狀 會選擇㈣雜節點。尤其是在第一子樹 二從終極根節點™叩開始，利用向下移動來決 達的葉=====的=節_ ==例中’所有葉片節點的數量為二二:定ί 用第-級於所有葉片節點數量。然後，採 中該第一級節點的特徵在於，可到達的葉片節：4 13 201234290 所有葉片節點數量，其子節點之可到達的葉片節點數量都 =於所有葉片節點的數量；該第.二級節點的特徵在於，該 達的葉片節點數量及其至少—子節點都等於所 ^片卽點數量。這是因為，假設節點m為節點n的親代 即並且可從這兩個節點到達所有葉片節點，也就是節 為第二級節點’然後採用節點n當成根節點的子樹狀 必須為採用節點m當成根節點的子樹狀結構Μ之 Ιί=相較於樹狀結構Μ ’樹狀結構Ν必須内含較 點當成適當根節點，並域該移除。#慮將第一級即 $實線方塊指出部分節點的可到達葉片 根節點™ng的可到達葉片節點及其兩 23之二I之數量為6’此外節點12的一個子節點 知伙二Α ^、、片卽點數量也是卜如此’該根節點1^1^ =點2都為第二級節點’並且應該移除；而應該將節 所λ 當成候選根節點。如此，初步從圖4β内 ，料工、结構中選擇兩健€子樹狀結構，如圖4C 節點樹狀結構分顧節點11和節點23當成根 斷。刀5,定的子樹狀結構上執行進-步判 且選擇該等節點數量最少的該子樹目標Ϊ f8個 =且内=兩個子咖 ==，涉到12個節點，如此緒構⑺小於 J 郎點之間的關聯性要超過(1)内節點之 間的關聯性。因此，應該選定結構(2)當成目標結構，來反 201234290 映出葉片節點之間的階層資訊。 雖然上面透過兩個步驟選定更小型的子樹狀結構當 成目標樹狀結構，吾人了解也可用其他方式來分析並選擇 子树狀結構。例如在一個具體實施例内，針對每一潛在子 樹狀結構，直接決定其内含的節點數量，並且選擇該等節 點數量最少的該子樹狀結構當成該目標樹狀結構。在其他 具體實施例内，首先選擇指定的葉片節點，然後針對每一 潛在子樹狀結構，決定從該根節點到該指定葉片節點的路 徑長度，也就是階層數量，並且選擇具有較少階層數量的 子樹狀結構當成目標樹狀結構。此方式可用來初步篩選子 樹狀結構，直接決定目標樹狀結構，或與節點數量判斷結 合’決定最終目標樹狀結構。 運用上述許多方法，從複數個子樹狀結構當中，可發 見相對小型的樹狀結構當成該目標樹狀結構。再者在一個 ^體實施例内，進一步分析並調整每一葉片節點在該目標 構内所在的階層，讓最終階層樹在結構内更對稱與均 尤其是，請參閱圖4C内的結構(2)，由於結構(2)較小， 所以選定當成該目標樹狀結構。不過在此樹狀結構内，雖 然全都是葉片節點，不過節點F和其他葉片節點A-E並不 ，同階層上，換言之，從該根節點到每一葉片節點的路 乜長度並不相同。如此，該樹狀結構並非均衡式樹狀結 構。因為均衡式樹狀結構更有利於後續階層資訊分析，所 以可調整該目標樹狀結構讓它「均衡」。在一個範例中’ 針對較鬲階層上的節點F，也就是與該根節點的距離比其 15 201234290201234290 VI. Description of the Invention: [Technical Field of the Invention] The present invention relates to the field of business intelligence, and more particularly to a method and apparatus for obtaining class information of a flat material. [Prior Art] In recent years, Business Intelligence (BI) technology provides comprehensive business information related services, such as performing data analysis, implementing data collection, generating data reports, and revealing data rules. Use = to analyze data and get reports to help companies make business decisions more efficiently. In business intelligence technology, the dimensioning and stratification of data is the basis for subsequent data analysis using stereo models. An example of a three-dimensional model of a circle and a multi-level data. In this example, the data on product sales is spread over three axes, namely time (X-axis), location (y-axis), and product (2-axis). The organization is organized to describe the sales volume and time, location, and Functional relationship between products. Further, the structure of the sales volume data is divided into a plurality of levels by each dimension, so that the data analysis and management can be performed at the root and layer. For example, in the geography dimension, the structure of the data in the sales is divided into the sales volume on each continent; for each piece of land, the sales volume is further divided into each country; for each country, according to the demand, the provinces and cities are used. distinguish. _ Ground, for the time, its structure is divided into year, season, month, and day according to demand: for production. Step-by-step according to the product category, series, model, etc. to the second & Kangting some dimensionalization and stratification of the data, using the stereo module ^ (10)P (On Line Analysis p followed by ss, online analysis department = = And the operation 'so presents the integration information from each dimension to the 201234290 class according to the user's needs. From the above examples, it can be understood that the dimensioning and stratification of the data provides significant convenience for data modeling and analysis of business intelligence. In addition to the general hierarchical enterprise data, we further want to apply the analysis and operation methods in business intelligence to other materials. However, in many fields, such as in the clinical field, the data is still organized in a "flat" manner. There are no examples of existing clinical data. In the example of Figure 2, typical examples of clinical data such as e-health include many types of information, such as the patient's main symptoms, diagnosis, treatment, etc. We can understand that All of this information is intensively arranged in a flat manner using clinical terms, and does not show data and data. The relationship between class information, and this is just the basis of intelligence analysis as a three-dimensional (four) and 〇LAp operation. Many planes of f data also exist in other commercial materials, due to the lack of class fj 'using existing intelligence modeling It is difficult to analyze and manage this kind of walking step with the operation method. This limits the systemization and information of the data. 2 I want to perform processing on the existing flat data to get the d layer Beixun' help in the plane. Subsequent analysis and management. [Summary of the invention] The problem of obtaining the information of the level of the plane data is given by the method of the invention of the invention. The ancestor 5 maps from the plane data—the same data i-tree project' to at least one node formed by the structuring terminology system, and obtains at least one subtree structure of the 201234290 within the above tree structure, the at least Each of the sub-tree structures adopts the at least one node as all of the blade nodes; and selects a target tree structure from the at least one sub-tree structure; Obtaining hierarchical information within the target tree structure. According to a second aspect of the present invention, there is provided an apparatus for obtaining hierarchical information of planar data, comprising: a node mapping unit configured to receive a data from the plane At least one data item of the same data set is mapped to at least one node in a tree structure formed by a structured term system; a sub-structure obtaining unit configured to obtain at least one sub-tree within the tree structure a structure, each of the at least one subtree structure using the at least one node as all of its blade nodes; a target structure selection unit configured to select a target tree structure from the at least one subtree structure And a hierarchical information obtaining unit 'set to obtain hierarchical information within the target tree structure. Using the method and apparatus of the present invention, hierarchical information between data items can be obtained from the planar organization data, thus helping subsequent planes Perform analysis and management on the data. [Embodiment] The present specification will explain the specific embodiments of the present invention by way of example and detailed examples. It is to be understood that the illustrated examples are illustrative only and should not be considered as limiting. As described above, the present invention provides such a method and apparatus for obtaining hierarchical information of planar materials. However, the information itself only contains a plurality of data items organized by the flat side 201234290, and does not know the relationship between each data item. The plurality of data items are usually recorded in the form of terms in the data. Inside. Therefore, to obtain the hierarchical information of the planar data, it is necessary to have the help of an external structured terminology system. The structured terminology system should record the canonical vocabulary in the field to which the data belongs, and organize the vocabulary in a hierarchical form to indicate many words. Category and affiliation. The following is a description of specific embodiments of the invention, using clinical data and structured terminology systems in the clinical field as examples. For the choice of clinical terminology systems, Systematized Terminology system of Medicine (SNOMED) is a widely used terminology system that provides a systematic organization of computer-processable collections, such as diseases, covering medical terminology in most clinical information fields. , discovery, surgery, microbes, medicines, etc. This enables a consistent approach to download, store, and collect clinical information for 10 times in professional and healthcare units. It also helps organize the content of medical records, reducing the data acquisition, coding methods, and the differences between clinical care for patients and research. e In other words, SN〇MED covers more than 365,000 clinical vocabulary and is assigned a unique “digital code, only-name” (for i “complete 19 and one description”. The above multiple words are organized into drugs. , including the vocabulary class of the clinical program, and the hierarchical class have many classification sub-levels, for example: medicine i 曰, kBa bed dysfunction related words can be classified according to body parts, etiology, etc. 201234290, thus obtaining further classification Hierarchy. - Different vocabulary within a class or across classes uses about 1 and is a "relationship". So SNOMED is a _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Considering the "slave" relationship between the capsules, a term relationship diagram with a tree structure can be obtained, where each term is a node of the tree structure; and the connecting lines between the nodes in the figure represent between the Lang points The affiliation. Without losing generality, it can be assumed that there is one of the most common concepts used as the root node of all vocabulary. Therefore, the node mosquito is "Thing", so all the festivals "Things connected to the root node" become their child nodes. As mentioned above, since the classification can be performed between vocabularies from different points of view, each node has multiple child nodes and multiple parent nodes (parent) N〇(jes). According to the characteristics of SN0MED mentioned above, it is better to use SN〇MED as a structured terminology system to describe the hierarchical relationship between clinical vocabulary. However, we understand that the choice of clinical terminology system is not limited to SN〇meD. It is also possible to use normalized and structured terminology systems that have been developed or will be developed in the future, such as the MedDRA terminology system. This terminology system can form a tree structure from different viewpoints and different aspects, thus expressing the relationship between nodes representing vocabulary. As for data in other fields, such as data on biological species, data in the chemical field, etc., there is also a corresponding structured terminology system. As described above, these structured terminology systems can organize standard vocabulary in the field into a tree structure. In order to explain in detail the 'bottom combination of representative clinical data and SNOMED terminology system' to illustrate the present invention Embodiments 201234290 Figure 3 shows a flow chart of a method in accordance with an embodiment of the present invention. As shown in Figure 3, a method for obtaining hierarchical information of planar data in accordance with an embodiment of the present invention includes: Step 31, wherein At least one data item of the same data set in the plane data is mapped to at least one node in a tree structure formed by a structured term system; and in step 32, at least one sub-tree structure in the tree structure is obtained, the at least Each of the subtree structures employs the at least one node as all of its blade nodes; in step 33, a target tree structure is selected from the at least one subtree structure; and in step 34, the target tree structure is obtained Class information within. In particular, in step 31, the data items in the planar data are located to the tree structure formed by the structured terminology system. Thus, the data set is first taken from the plane data, so that a plurality of data items in the data set are obtained, so that the data items from the same data set are analyzed, and the information from the same dimension is reflected. For example, in the clinical data shown in Figure 2, which uses electronic medical records as an example, each vertical column is treated as a data set from one dimension that reflects medical record information. In particular, each data item in the data set of the second vertical column is used to describe the main symptoms of the case, and each data item in the data set of the seventh vertical column is used to explain the diagnosis conclusion of the case, and the eighth vertical block. Each data item in the data set is used to describe the treatment of the case. Therefore, data items from the same vertical column (that is, the same data set) are used as targets for subsequent step analysis. Next, in order to obtain a plurality of data items, each data item is mapped to a vocabulary within the structured term system. In one embodiment, the planar data is clinical data and the structured terminology system is the SNOMED terminology system described above. Many clinical data currently use the standard terminology in the 201234290 SNOMED terminology system to record clinical information, and some even use the code in the SNOMED terminology system to record and store data. In this case, the vocabulary mapping of the data items in the clinical data to the SNOMED terminology system can be achieved by using only the vocabulary or code search and pairing. In the case where the clinical data is not recorded in a standard vocabulary, string pairing and fuzzy logical pairing between the data item and the vocabulary can be additionally performed. In some embodiments, the explanation or explanation of the vocabulary within the terminology system is sometimes referred to as assistance. For the flat data of other content, if the same flat data has been recorded in the vocabulary or code in the structured terminology system, the search and pairing is performed on the vocabulary or code to directly map the data item to the vocabulary; Fuzzy logic pairing can be additionally performed without being recorded in the canonical vocabulary. In addition, there are many ways to match vocabulary matches in the industry, and those skilled in the art can choose the appropriate method to perform mapping and mapping of data items and vocabulary. As such, each acquired data item is mapped to a vocabulary within the structured terminology system. In addition, because, as described above, since the structured terminology system is organized according to hierarchical organization, thereby forming a tree structure of terms, the vocabulary corresponding to the individual material items is used as a node within the tree structure. Therefore, the data item has been located within the tree structure. Figure 4A shows a tree structure diagram in accordance with an embodiment of the present invention. This configuration shows a part of the tree structure formed by the SN〇MED term system, where each node corresponds to a vocabulary, the connecting line with arrows shows the parent-child relationship between the nodes' and the "Thing" is the entire tree structure. Root node. Through the mapping in step 31 above, the data item is located to a specific node within the tree structure. In Figure 4A, the nodes from the data item map 201234290 are displayed as nodes eight, 6, 〇0, mound, and 1?. Next, the method of the present embodiment proceeds to step 32 where it is found that the node mapped from the data item is used as at least one subtree structure of all of its leaf nodes. Still referring to Fig. 4A, the process of step 32 is to find at least one subtree structure within the entire tree structure, each of which uses nodes A-F as all of its blade nodes. The person who wants to determine the candidate subtree structure needs to apply the connection relationship between the nodes in the tree structure. In one embodiment, a structured terminology system (e.g., SNOMED) that forms the tree structure is launched in the form of linked 〇pen data'LOD. In this form, the relationships between the nodes within the tree structure are all described and stored in the format of RDF triples. As the skilled person is familiar with, 'RDF triple uses < subject, predicate, and vocabulary> to express many meanings and relationships. The affiliation of nodes A and B (or parent-child relationship) can be represented by RDF triple as <n〇deA,subClassOf, nodeB>. For semantic languages, there is a concept of owl:Thing in the data; each individual item in the LOD data is its member, or its child node. Therefore, to query the parent node of the node "childNode" in the LOD, the following SPARQL query can be used: Select ?parentNode where {?parentNode rdfs:subClassOf <childNode>}, thus obtaining the value of the parent node. A child node of a known parent node can also be queried in a similar manner. In this case, the parent-child relationship between the nodes can be easily obtained through the core term subClassOf. In other embodiments, the structured terminology system is stored in other specified formats. Therefore, 'using the affiliation written in other specified formats 201234290, can obtain the parent-child relationship between the nodes in the ship's structure (four) point _ parent-child relationship, can be in the tree club: r moves up or down, so you can perform i through this shift: Zhao dimorphism: the root node of the structure is the root node, and the road is moved up: Shi::: I : 2 to 2 The above SMRQL query obtains at least the parent node with the "-C 2 0f" blade node; then 2 = ΡΪ · sequentially obtains the upper node with higher order, the single path of the point = 2: the blade Node to the common node between the root segments, such as: and; = = find two = subtree from the blade node to the root node Thing thereby obtaining a heart map 4β display according to a specific embodiment of the present invention, from The sub-tree structure obtained by the sheep in Fig. 4 is shown. As shown in FIG. 4A, the obtained tree structure belongs to a part of the tree structure of FIG. 4; the structure has: the blade nodes are nodes A_F mapped from the data item, and points! It is TMng. However, it can be found from the observation of this subtree structure of the root node two = slice node AF, which is not the only substructure of the node Α_ρ as a blade node, but also another subtree structure 'for example Use node U as the root node for another 12 201234290 a subtree structure. That is, by moving between the blade node AF and the ultimate root node Thing, multiple subtree structures can be determined, and from a particular viewpoint or aspect, these subtree structures can still reflect potential hierarchical relationships between nodes. . Therefore, one of the obtained multi-subtree structures can be determined as the target tree structure according to requirements, and the hierarchical tree structure of the node is reflected by the target tree structure, as shown in step 33. In a specific embodiment, in order to make the finally obtained hierarchical information more relevant, it is necessary to further screen the obtained multi-subtree structure, and select a relatively small "small" tree structure to reflect the hierarchical information, because In a hierarchical tree with a relatively "small" structure, nodes have a tighter correlation' that better reflects the specific classification and theme of the tree. The selection process described above will be described below in conjunction with the subtree structure example shown in Fig. 4B. In one embodiment, two steps are used to select from multiples. First, for the child age structure of the Zlhmg from the blade node surname shown in Figure 4B (the _ is $-subtree will select the (4) hybrid node. Especially in the first subtree 2 from the ultimate root node TM叩, Use the downward movement to determine the leaf ======section=== In the example, the number of all blade nodes is two or two: the number of all blade nodes is determined by the first level. Then, the number is selected. The first-order node is characterized by the reachable blade segments: 4 13 201234290 The number of all blade nodes, the number of reachable blade nodes of its child nodes = the number of all blade nodes; the second-order node is characterized by The number of blade nodes and their at least-child nodes are equal to the number of slices. This is because it is assumed that node m is the parent of node n and can reach all blade nodes from these two nodes, that is, the node. For the second-level node' then use the node n as the sub-tree of the root node must be the sub-tree structure using the node m as the root node Ι = = compared to the tree structure Μ 'tree structure Ν must contain Point as an appropriate root node, and the domain should be removed.# Considering that the first level, that is, the solid line, indicates the reachable blade node of the reachable blade root node TMng of the partial node and the number of two of the two 23's is 6', and a child node of the node 12 is known to be two. The number of slices is also such that 'the root node 1^1^ = point 2 is the second-level node' and should be removed; instead, the node λ should be regarded as the candidate root node. Thus, initially from Figure 4β In the material worker and structure, two tree structures are selected, as shown in Fig. 4C, the node tree structure divides node 11 and node 23 into roots. Knife 5, the determined subtree structure performs the step-by-step judgment and selects The subtree target with the least number of such nodes Ϊ f8 = and inner = two sub-cookies ==, involving 12 nodes, so the correlation between the imperative (7) and the J-lang points is more than (1) the inner node The correlation between the two. Therefore, the structure (2) should be selected as the target structure to reflect the hierarchical information between the blade nodes in 201234290. Although the smaller subtree structure is selected as the target tree structure through two steps above. I understand that there are other ways to analyze and select subtree knots. For example, in a specific embodiment, for each potential subtree structure, the number of nodes included therein is directly determined, and the subtree structure with the least number of such nodes is selected as the target tree structure. In an embodiment, the specified blade node is first selected, and then for each potential subtree structure, the path length from the root node to the specified blade node, that is, the number of levels, and the subtree having a smaller number of levels are selected. The structure is regarded as the target tree structure. This method can be used to initially screen the sub-tree structure, directly determine the target tree structure, or combine with the number of nodes to determine the final target tree structure. Using many of the above methods, from a plurality of sub-trees Among the structures, a relatively small tree structure can be seen as the target tree structure. Furthermore, in a physical embodiment, the hierarchy of each blade node within the target structure is further analyzed and adjusted, so that the final hierarchical tree is more symmetrical and especially in the structure, see the structure in FIG. 4C (2) Since the structure (2) is small, it is selected as the target tree structure. However, in this tree structure, although all are blade nodes, the node F and other blade nodes A-E are not, the same level, in other words, the length of the path from the root node to each blade node is not the same. Thus, the tree structure is not a balanced tree structure. Because the balanced tree structure is more conducive to subsequent class information analysis, the target tree structure can be adjusted to make it "balanced". In one example, 'for node F on the lower level, that is, the distance from the root node is 15 201234290
他葉片節點還要短的節點，設定假子節點 點F相同的内容，但是位於與其他葉片 其具有和fP 層上，如圖4D内所示。如此，該:點相同的階 有葉月節點都位於相同階層上，藉此實=樹狀結構内所 吾人了解，若葉片節點之間相差—卿内的均衡。 較高階層上設定三個以上階層的假子1 ’則需要在 :如此最終讓目標樹狀結構内所有“= 總結來說，透過上述方法，可獲得 射的節點當成葉片節點之小型並獲二=資= 構。根據此點，在步驟34上，從該目# : “樹狀結 節點之間的階層資訊’如此可知道對；苹片 項目間之關聯性。例如；透過圖4D内戶 =郎點的資料 構，可獲得葉片節點A-F之間的階層=的目標樹狀結 應至葉片節點A-F的臨床資料項“之既有=對 應至葉片節點A_C的資料項目屬於相 對乂：= ::::標樹”構中資訊藉：表 從該根節放=著 對f至該樹狀結構的階層表。圖4Ε顯示^應至圖4D = 該樹狀結構的階層表。在其他 心 内 也可組織為其他H ，該階層資訊 201234290 根據上述獲得的階層資訊，可在平面組織的資料項目 上執行於商業情報中廣泛調整過的〇LAP分析與操作，藉 此從分散並且平面的資料項目中揭露既有關聯性與資料 規則，如此在資訊上能執行更好的分析與管理。 根據相同的發明概念，本發明也提供一種獲得平面資 料的階層資訊之設備。圖5顯示根據本發明具體實施例的 s史備方塊圖。如圖5内所示，本發明具體實施例的設備5〇 包括：一節點映射單元51，其設置成將來自該平面資料内 了相同資料集的至少一資料項目，映射至由一結構化術語 系統所形成之一樹狀結構内的至少一節點；一子結構獲得 單元52’其設置成從該上述樹狀結構獲得至少一子樹狀結 構，該至少一子樹狀結構的每一者採用該至少一節點做為 其所有葉片節點；一目標結構選擇單元53，其設置成從該 至少二子樹狀結構當中選擇一目標樹狀結構；以及一階層 資訊獲得單元54,其設置成獲得該目標樹狀結構内的該階 層資訊。 尤其疋，節點映射單元51用於將平面資料内的資料 ^ ^疋位至該結構化術語系統所形成的樹狀結構。如此首 先=點映射單元51從該平面資料中擷取一資料集，並且 獲得該資料集内複數個資料項目，如此要分析的資料項目 來自相同的資料集，並且反映出相同維度的資訊。接下 ，二為了獲得複數個資料項目，節點映射單元51將每一 丄斗，目映射至該結構化術語系統内的辭彙。在已經用結 映化，n。系統内規範詞彙描述該平面資料的情況下，節點 元51在詞彙或代碼上僅執行搜尋與配對，就可實 貝料項目映射至詞彙。在平面資料並未用規範詞彙記錄 17 201234290 的情況下，節點映射單元51可另外執行資料項目與詞彙 之間的字串配對與模糊邏輯配對，藉此將資料項目映射至 5¾彙。進一步，因為該結構化術語系統根據階層來組織詞 彙，藉此形成術語的樹狀結構，其中一個詞彙為該樹狀結 構内的一個節點，當節點映射單元51將資料項目映射^ 詞彙時’該等資料項目同時映射至該樹狀結構内的節點。 接I來’子結構獲得單元52找出在上述樹狀結構内， 採用從資料項目映射的節點，當成其全部葉片節點的 一子樹狀結構。 為了獲得候選子樹狀結構，子結構獲得單元52 =節：之間許多種格式的連接關係(尤其是親子關係)之 存該結構化術語系統。若獲得節點之 則子結構獲得單元52可在該樹狀結構内 動’並且透過這種移動來決定子樹狀結構。 姓播ί —個具體#施_，子節點獲得單元52從該她 到達的路徑，其4算::動/1個別葉片節點上可 jh 、以等葉片郎點為由節點映射單元51铋 貝枓項目映射之節點。子紝椹平7° M從 其中所牽涉節點結合當爾軸 内，子社鮮子樹狀結構。在其他具體實施例 節點τΓ 從料料節點向上移動到_ 卸點Thmg，藉此形成從該尊 功抑根 徑。此後針對該獲得的多個路片找即==節點的路 =到該根郎點Thing的—第一子樹 以第—子樹狀結構實際上包括許多可能的子樹狀、=:所 201234290 以可依需求進-步筛選所獲得的多個 選擇-個適當的子樹狀結構當 構，從其中 出階層資訊。 不柄狀、，吉構，如此反映 分析由子結構獲得單元52 並從其中選擇能夠反映出 然後目標結構選擇單元53 所獲得的該等多個子樹狀結構， 節點階層資訊的目標樹狀結構。 你1因兵媸貫苑例内，目標結構 個步驟，來分析多個顿狀結構，53運用兩 構當成該目標樹狀結構。首先在該第樹狀結 該終極根節點Thillg開始向下移動 ’構内’從 ，片節點數量。然後，採用第'級節可到 點’並且移除第二級節點，其中該第—級：，選根節 可到達的料節點數量等於所有料徵在 第一級節點的子節點中可到達的葉 數里，並且 所有葉片節點數量；該第二級節點的特徵在量全都小於 點可到達的葉片節點數量及其至少—；，第二級節 片節點數量。 、^ ?㈣等於所有葉 接下來，在初步選定的子樹狀結構上 ％ 斷，其中已經移除該第二級節點。尤其是，仃進一步判 單元53決定每一子樹狀結構内含的節點數量標、％構選擇 該等節點數量最少的該子樹狀結構當成該目^妹·^且選擇 透過上述個別單元，設備50可獲得多個子 其採用從資料項目映射的節點當成葉片節點，並丨狀結構， 中更小的樹狀結構當成該目標結構。進—·丄、’找出其 Y在一個具體實 19 201234290 備5G也包括—均衡單元(未顯示），設置成分析 並調整葉片節點在該目標結構内所在的階層讓最終目標 結構^對稱與均衡。尤其是，若該目標結構内的該等個別 $片節點位於不同階層上’職均衡單元可利用設定假子 節點來均衡該目標結構，如此該目標樹狀結構内的所有葉 片節點都在相同階層上。 ' 一在已經決定目標樹狀結構的基礎上，階層資訊獲得單 兀』4從該目標樹狀結構中掘取該階層資訊，藉此顯示個 =卽點之_關雜，並且進—倾示對應別 資料項目間之階層資訊。 根據本發明具體實施例獲得平面資料的階層資訊之 設備50的詳細範例類似於上述方法之範例，因二為:^簡 化而省略其細節。 運用許多具體實施例的方法及設備’藉由結構化術語 系統可獲得平面資料的階層資訊，如此幫助在平面 進行後續分析與管理。 精，技術人士可了解’使用電腦可執行指令及/或包含 在處理H㈣碼内，可實細於麟平面⑽的階層資訊 之上述方法及設備，其中該控制碼提供於，例如磁碟、ceD 或dvd_rgm賴承載、例如唯讀記賴這類可程式 記憶體或勤絲或電錢紐辆㈣細。本發明且 體實施例的設備及其組件都可由硬體電路實施，例如大2 積體電路或閘極陣列、例如邏輯晶片或電晶體的半導體’ 或例如場效應可程式閘極陣列、可程式邏輯裝置的可程式 20 201234290 硬體裝置，或可由許多種處理器執行的軟體來實施，或可 由上述硬體電路與軟體的組合來實施。執行本發明操作的 軟體與電腦程式碼可用任何-❹種程式語言的組合來 撰寫，包括但不受限於例如java、Sma丨ltalk、c++等的物 ^導向程^語f ’錢例如「c」程式語言絲似程式語 =傳統程序程式語言。該程式碼可在本地或遠端電腦上 執行’以達成所要的操作。 雖然已麟合詳_具體實齡j，詳述域本發 :=平：資料的階層資訊之方法及設備，本發明並不受 =發===領域之下’精通技術人士 乂㈣L 進仃許多變化、替換以及改變。吾 ’所有· '替換與改變仍舊位於本發明中由 後附申請專利範圍所保護的領域内。 千由 【圖式簡單說明】 =Πί與多階層資料的立體模型之範例； 圖2,、.員不現有臨床資料的範例； 結構； 根據，明具體實施例的方法之流程圖； 顯雜據本發明具體實關的触結構圖； .至圖4D顯不根據本發明具體實施例的子樹狀 及 圖狃顯示對應至圖4D内該樹狀結構的階層表； 圖5顯示根據本㈣具體實施_設備方塊圖。 以 201234290 【主要元件符號說明】 50 設備 51 節點映射單元 52 子結構獲得單元 53 目標結構選擇單元 54 階層資訊獲得單元 22The node of the blade node is also short, setting the same content of the fake child node F, but it is located on the fP layer with other blades, as shown in Figure 4D. In this way, the points with the same order are all located on the same level, so that the real person in the tree structure knows that if the leaf nodes are different, the equilibrium within the Qing. At the higher level, setting more than three levels of the fake 1' needs to be: so that all the targets in the target tree structure are finally "= In summary, through the above method, the node that can be obtained as the blade node is small and won two. According to this point, in step 34, from the item #: "the hierarchical information between the tree nodes", it is known that the pair; the correlation between the items of the film. For example, through the data structure of the household=lang point in Fig. 4D, the target tree-like node of the level== between the leaf nodes AF can be obtained. The clinical data item of the leaf node AF “is both = corresponding to the leaf node A_C The project belongs to the relative 乂:= :::: standard tree structure in the information borrowing: the table from the root node = the pair of f to the tree structure of the hierarchical table. Figure 4 shows the hierarchical table of the tree structure shown in Figure 4D. In other hearts, it can also be organized into other H. The class information 201234290 can be performed on the data items of the plane organization based on the extensively adjusted 〇LAP analysis and operation in the business information. The flat data project exposes both the relevance and the data rules, so that better analysis and management can be performed on the information. According to the same inventive concept, the present invention also provides an apparatus for obtaining hierarchical information of planar materials. Figure 5 shows a block diagram of a s history in accordance with an embodiment of the present invention. As shown in FIG. 5, the device 5 of the embodiment of the present invention includes: a node mapping unit 51 configured to map at least one data item from the same data set in the plane data to a structured term Forming at least one node in the tree structure; a substructure obtaining unit 52' is configured to obtain at least one subtree structure from the tree structure, each of the at least one subtree structure adopting the At least one node is used as all of its blade nodes; a target structure selecting unit 53 is provided to select a target tree structure from the at least two subtree structures; and a hierarchical information obtaining unit 54 is provided to obtain the target tree This level of information within the structure. In particular, the node mapping unit 51 is configured to clamp the data in the plane data to the tree structure formed by the structured term system. Thus, the first-point mapping unit 51 extracts a data set from the plane data, and obtains a plurality of data items in the data set, so that the data items to be analyzed are from the same data set, and reflect the information of the same dimension. Next, in order to obtain a plurality of data items, the node mapping unit 51 maps each of the buckets to a vocabulary within the structured terminology system. In the already used, n. In the case where the standard vocabulary in the system describes the plane data, the node element 51 performs only the search and pairing on the vocabulary or code, and the real item item is mapped to the vocabulary. In the case where the plane material is not in the canonical vocabulary record 17 201234290, the node mapping unit 51 can additionally perform string pairing and fuzzy logic pairing between the data item and the vocabulary, thereby mapping the data item to the sputum. Further, because the structured terminology system organizes vocabulary according to the hierarchy, thereby forming a tree structure of terms, wherein one vocabulary is a node within the tree structure, and when the node mapping unit 51 maps the data item to the vocabulary, The data items are simultaneously mapped to nodes within the tree structure. The sub-structure obtaining unit 52 finds that within the above tree structure, the node mapped from the data item is used as a sub-tree structure of all the blade nodes. In order to obtain the candidate subtree structure, the substructure obtaining unit 52 = section: a plurality of formats of connection relationships (especially parent-child relationships) between the structured terminology systems. If a node is obtained, the substructure obtaining unit 52 can move within the tree structure and determine the subtree structure by such movement. The last name broadcasts a specific #施_, the child node obtaining unit 52 from the path that she arrives, the 4 counts:: move / 1 individual blade nodes can be jh, with the equal blade point as the node mapping unit 51 mussel节点The node of the project map. Ziyuping 7° M from which the nodes involved are combined with the Dang axis, the sub-tree is a tree structure. In other embodiments, the node τ 向上 moves up from the stock node to the _ unload point Thmg, thereby forming a root from the dignity. Then, for the obtained plurality of road slices, find the road of == node = to the root point of the Thing - the first subtree with the first subtree structure actually includes many possible subtrees, =: 201234290 The plurality of choices obtained by the step-by-step screening can be selected as needed - an appropriate sub-tree structure is constructed, from which the hierarchical information is derived. The object structure is obtained by the sub-structure obtaining unit 52 and the target tree structure capable of reflecting the plurality of sub-tree structures and node level information obtained by the target structure selecting unit 53 is selected therefrom. You 1 analyze the multiple structures in the example of the target, and use the two structures as the target tree structure. First in the first tree, the ultimate root node, Thillg, begins to move downwards. Then, the first stage can be used to the point and the second level node is removed, wherein the level: the number of material nodes reachable by the selected root node is equal to all the elements reachable in the child nodes of the first level node The number of leaves, and the number of all blade nodes; the characteristics of the second-level nodes are all smaller than the number of blade nodes reachable by the point and at least -; the number of second-stage node nodes. , ^ (4) equals all leaves Next, % breaks on the initially selected subtree, where the second level node has been removed. In particular, the further determining unit 53 determines the number of nodes included in each subtree structure, and selects the subtree structure with the smallest number of such nodes as the target and selects to pass through the individual units. The device 50 can obtain a plurality of sub-nodes that are mapped from the data item as blade nodes, and a smaller tree-like structure as the target structure. Into - 丄, 'find its Y in a concrete real 19 201234290 备 5G also includes - equalization unit (not shown), set to analyze and adjust the level of the blade node in the target structure to make the final target structure ^ symmetry balanced. In particular, if the individual $chip nodes in the target structure are located at different levels, the 'employment equalization unit can use the set hypothetical node to equalize the target structure, so that all the blade nodes in the target tree structure are in the same hierarchy. on. 'After the target tree structure has been determined, the class information is obtained from the single tree』 4 from the target tree structure to extract the class information, thereby displaying the = 卽 之 、 、 、 、 、 、 、 Corresponding to the class information between other data items. A detailed example of the device 50 for obtaining hierarchical information of planar data according to an embodiment of the present invention is similar to the above-described method, and the second is: simplification and omitting details thereof. The method and apparatus of many specific embodiments are utilized to obtain hierarchical information of planar data by a structured terminology system, which facilitates subsequent analysis and management in the plane. Fine, technical personnel can understand the above-mentioned methods and devices for using computer-executable instructions and/or class information that can be subdivided into a lining plane (10) in a H (four) code, wherein the control code is provided, for example, on a disk, ceD Or dvd_rgm depends on the load, such as reading only such programmable memory or wire or electric money (four) fine. The apparatus of the present invention and its components can be implemented by a hardware circuit, such as a large integrated circuit or a gate array, a semiconductor such as a logic chip or a transistor, or a field effect programmable gate array, or a programmable Programmable 20 201234290 hardware device, or software that can be executed by a variety of processors, or can be implemented by a combination of the above-described hardware circuit and software. The software and computer code for performing the operations of the present invention can be written in any combination of any programming language, including but not limited to, for example, java, Sma丨ltalk, c++, etc. "Program language is like a program language = traditional programming language. This code can be executed on a local or remote computer to achieve the desired operation. Although it has been detailed _ concrete age j, detail domain domain: = Ping: data of the class information method and equipment, the invention is not = = = = = under the field 'proficient in technology people 乂 (four) L Many changes, substitutions, and changes. The 'all' replacement and alterations are still within the scope of the invention as protected by the scope of the appended claims. An example of a three-dimensional model of a simple schema and a multi-level data; Figure 2, an example of an existing clinical data; a structure; a flow chart of a method according to a specific embodiment; The touch structure diagram of the present invention is specifically implemented; the subtree and the diagram showing the tree structure corresponding to the tree structure in FIG. 4D are not shown in FIG. 4D; FIG. 5 shows the concrete structure according to the present (4). Implement the _device block diagram. 201234290 [Description of main component symbols] 50 Device 51 Node mapping unit 52 Sub-structure obtaining unit 53 Target structure selecting unit 54 Hierarchy information obtaining unit 22