TWI696923B - Recommendation system for patentable prediction and invalid comparison and analysis method thereof - Google Patents
Recommendation system for patentable prediction and invalid comparison and analysis method thereof Download PDFInfo
- Publication number
- TWI696923B TWI696923B TW107121551A TW107121551A TWI696923B TW I696923 B TWI696923 B TW I696923B TW 107121551 A TW107121551 A TW 107121551A TW 107121551 A TW107121551 A TW 107121551A TW I696923 B TWI696923 B TW I696923B
- Authority
- TW
- Taiwan
- Prior art keywords
- association rule
- search
- association
- strength
- database
- Prior art date
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
本發明涉及一種推薦系統及其分析方法,特別是根據專利分類號建立關聯規則並獲得相應的關聯規則強度,用以產生可專利預測及專利無效比對推論之建議。The present invention relates to a recommendation system and its analysis method, in particular to establish association rules based on patent classification numbers and obtain corresponding strengths of association rules to generate recommendations for patent prediction and patent invalidation comparison inference.
近年來,隨著智慧財產權的普及與蓬勃發展,各種以專利資料庫為基礎的相關應用便如雨後春筍般出現,如:專利地圖分析、專利資料探勘及專利鑑價等等。In recent years, with the popularization and vigorous development of intellectual property rights, various related applications based on patent databases have sprung up, such as: patent map analysis, patent data exploration and patent valuation, etc.
一般而言,傳統專利資料庫的運用大多朝向巨量分析的視覺化、機器學習、深度學習、語意分析的研究方向發展。然而,對於大數據的資料探勘來說,資訊科學在專利資料庫的運用上,大多朝向企業併購與智慧資本的高度需求來加以呈現,鮮少關注在研發體系的實質應用。另一方面,視覺化軟體呈現專利數據的美化與互動性,往往對研發人員的參考意義不大。再者,企業智慧財產權(Intellectual Property, IP)的發展大多由法務領域的人員主導與管理,針對其領域的需求大多限制在專利檢索的比對性,所以專利分析跨越至研發人員的需求往往無法在一般的企業中被彰顯出來,因此,也限制了專利分析的完整發展,導致研發人員無法將專利分析融入開發的工作流程中,例如:無法從專利分析中得到組合不同技術的點子,或是在遭遇競爭對手的專利阻礙時,無法獲得用於無效比對推論的前案建議,故具有可專利預測及無效比對不便之問題。Generally speaking, the use of traditional patent databases mostly develops toward the research direction of visualization of massive analysis, machine learning, deep learning, and semantic analysis. However, for data exploration of big data, the application of patent science database in information science is mostly presented towards the high demand of enterprise mergers and acquisitions and smart capital, and little attention is paid to the actual application in the research and development system. On the other hand, visual software presents the beautification and interactivity of patent data, which is often of little significance to R&D personnel. In addition, the development of Intellectual Property (IP) in enterprises is mostly dominated and managed by personnel in the legal field, and the demand for its field is mostly limited to the comparison of patent search, so patent analysis can not span the needs of R&D personnel. It is highlighted in general companies, and therefore, limits the complete development of patent analysis, resulting in the inability of R&D personnel to integrate patent analysis into the development workflow. For example, it is impossible to obtain ideas for combining different technologies from patent analysis, or When encountering patent obstacles of competitors, the previous case proposal for invalidation inference cannot be obtained, so there is the problem of inconvenience in patent prediction and invalidation comparison.
有鑑於此,便有廠商提出應用人工智慧建立技術功效矩陣圖的技術,提供研發者了解技術聚集點以及技術空白點,進而規避技術熱點而發現新的研發方向。然而,此一方式需要耗費大量的計算機運算能力,而且無法呈現不同技術的結合可能性及關聯性,所以容易導致研發者在單一技術手段中鑽牛角尖,對研發者而言獲得的幫助十分有限,難以直接根據技術功效矩陣圖來發想出具可專利性的技術,或是做為專利無效推論的論證基礎,因此仍然無法有效解決可專利預測及無效比對不便的問題。In view of this, some manufacturers have proposed the technology of using artificial intelligence to establish a technology efficacy matrix diagram to provide developers with an understanding of technology aggregation points and technology gaps, so as to evade technology hot spots and discover new research and development directions. However, this method requires a lot of computer computing power and cannot present the possibility and relevance of the combination of different technologies, so it is easy to cause the developer to drill the tip in a single technical means. The help for the developer is very limited and difficult Directly inventing patentable technology based on the technical efficacy matrix diagram or as the basis for the demonstration of patent invalidation inferences, it still cannot effectively solve the inconvenience of patentable prediction and invalidation comparison.
綜上所述,可知先前技術中長期以來一直存在可專利預測及無效比對不便之問題,因此實有必要提出改進的技術手段,來解決此一問題。In summary, it can be seen that the prior art has had problems of inconvenience in patent predictability and invalidity comparison for a long time, so it is necessary to propose improved technical means to solve this problem.
本發明揭露一種可專利預測及無效比對之推薦系統及其分析方法。The present invention discloses a patentable prediction and invalid comparison recommendation system and its analysis method.
首先,本發明揭露一種可專利預測之推薦系統,此系統包含:專利資料庫、檢索模組、分析模組及處理模組。其中,專利資料庫用以儲存專利文件,每一專利文件皆包含專利分類號;檢索模組用以提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫以進行專利檢索,查詢出符合檢索條件的專利文件;分析模組用以載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度;處理模組用以將關聯規則強度為弱的關聯規則中的專利分類號進行組合以輸出為衍生專利建議。First, the present invention discloses a patent predictable recommendation system. The system includes: a patent database, a search module, an analysis module, and a processing module. Among them, the patent database is used to store patent documents, and each patent document contains a patent classification number; the search module is used to provide key search conditions, and the key search conditions are sent to the patent database for patent search, and the search finds out Patent documents for search conditions; the analysis module is used to load the searched patent documents, and analyze the patent classification number of the loaded patent documents with an association rule algorithm, and establish association rules according to the analysis results, each association rule contains at least Two patent classification numbers and one association rule strength; the processing module is used to combine the patent classification numbers in the association rules with weak association rule strength to output as derivative patent suggestions.
另外,本發明揭露一種可專利預測之分析方法,其步驟包括:在專利資料庫中儲存專利文件,每一專利文件皆包含專利分類號;提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫以進行專利檢索,查詢出符合檢索條件的專利文件;載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度;以及將關聯規則強度為弱的關聯規則中的專利分類號進行組合以輸出為衍生專利建議。In addition, the present invention discloses a patent predictable analysis method. The steps include: storing patent documents in a patent database, each of which contains a patent classification number; providing key search conditions, and transmitting the key search conditions to the patent The database is used for patent search to search for patent documents that meet the search conditions; load the searched patent documents, and analyze the patent classification number of the loaded patent documents with an association rule algorithm, and establish association rules based on the analysis results. An association rule includes at least two patent classification numbers and one association rule strength; and the patent classification numbers in the association rules with weak association rule strength are combined to output as derivative patent suggestions.
接著,本發明揭露一種專利無效比對之推薦系統,此系統包含:專利資料庫、檢索模組、分析模組及處理模組。其中,專利資料庫用以儲存專利文件,每一專利文件皆包含專利分類號;檢索模組用以提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫以進行專利檢索,查詢出符合檢索條件的專利文件;分析模組用以載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度;處理模組用以將關聯規則強度為強的關聯規則中的專利分類號進行組合以輸出為專利無效推論建議。Next, the present invention discloses a recommendation system for comparing invalid patents. The system includes: a patent database, a search module, an analysis module, and a processing module. Among them, the patent database is used to store patent documents, and each patent document contains a patent classification number; the search module is used to provide key search conditions, and the key search conditions are sent to the patent database for patent search, and the search finds out Patent documents for search conditions; the analysis module is used to load the searched patent documents, and analyze the patent classification number of the loaded patent documents with an association rule algorithm, and establish association rules based on the analysis results. Each association rule contains at least Two patent classification numbers and one association rule strength; the processing module is used to combine the patent classification numbers in the association rules whose association rule strength is strong to output as patent invalidation inference suggestions.
接下來,本發明揭露一種專利無效比對之分析方法,其步驟包括:在專利資料庫中儲存專利文件,每一專利文件皆包含專利分類號;提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫以進行專利檢索,查詢出符合檢索條件的專利文件;載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度;以及將關聯規則強度為強的關聯規則中的專利分類號進行組合以輸出為專利無效推論建議。Next, the present invention discloses an analysis method of patent invalidation comparison. The steps include: storing patent documents in a patent database, each of which contains a patent classification number; providing key search conditions, and transmitting the key search conditions Go to the patent database for patent search, search for the patent documents that meet the search conditions; load the searched patent documents, and analyze the patent classification number of the loaded patent documents with the association rule algorithm, and establish association rules based on the analysis results Each association rule contains at least two patent classification numbers and one association rule strength; and the patent classification numbers in the association rules with strong association rule strength are combined to be output as a patent invalidation inference suggestion.
本發明所揭露之系統與方法如上,與先前技術的差異在於本發明是透過載入與檢索條件相符的專利文件,並且以關聯規則演算法對載入的專利文件進行分析,用以建立包含專利分類號及關聯規則強度的關聯規則,以及從關聯規則強度為弱/強的關聯規則中,將其包含的專利分類號進行組合以輸出成為衍生專利建議/專利無效推論建議。The system and method disclosed by the present invention are as above, and the difference from the prior art is that the present invention is to load patent documents matching the search conditions, and analyze the loaded patent documents with an association rule algorithm to create a patent The association rules of the classification number and the strength of the association rules, and from the association rules whose strength of the association rules are weak/strong, the patent classification numbers contained in them are combined to be output as derivative patent suggestions/invalidation inference suggestions.
透過上述的技術手段,本發明可以達成提高可專利預測及無效比對的便利性之技術功效。Through the above-mentioned technical means, the present invention can achieve the technical effect of improving the convenience of patentable prediction and invalid comparison.
以下將配合圖式及實施例來詳細說明本發明之實施方式,藉此對本發明如何應用技術手段來解決技術問題並達成技術功效的實現過程能充分理解並據以實施。The embodiments of the present invention will be described in detail below in conjunction with the drawings and examples, so as to fully understand and implement the implementation process of how the present invention uses technical means to solve technical problems and achieve technical effects.
在說明本發明所揭露之可專利預測及無效比對之推薦系統及其分析方法之前,先對本發明所自行定義的名詞作說明,本發明所述的關聯規則強度是指在同一關聯規則中的相關聯元素(即:專利分類號)彼此之間的連結強弱,如:強連結或弱連結,舉例來說,當這些元素頻繁出現的次數大於某一預設值可代表關聯規則強度為強,或稱之為強連結;反之則代表關聯規則強度為弱,或稱之為弱連結。在資料探勘的領域之中,關聯規則(Association Rule)分析是最常被使用的方法,其方法則大致是『if 前項antecedent(s) then後項consequent(s) 』的概念,目的在於找出資料庫中資料間彼此的關聯性。Before describing the patentable prediction and invalidity comparison recommendation system and analysis method disclosed by the present invention, the nouns defined by the present invention will be described first. The strength of the association rules in the present invention refers to the same association rules. The strength of the links between related elements (ie, patent classification numbers), such as strong links or weak links, for example, when these elements frequently occur more than a certain preset value may indicate that the strength of the association rule is strong, Or referred to as strong links; otherwise, the strength of the association rules is weak, or called weak links. In the field of data exploration, association rule (Association Rule) analysis is the most commonly used method, and its method is roughly the concept of "if antecedent(s) then afterconsequent(s)", the purpose is to find out The relevance of the data in the database.
以下配合圖式對本發明可專利預測及無效比對之推薦系統及其分析方法做進一步說明,請先參閱「第1圖」,「第1圖」為本發明可專利預測之推薦系統的系統方塊圖,此系統包含:專利資料庫110、檢索模組120、分析模組130及處理模組140。其中,專利資料庫110用以儲存專利文件,每一專利文件皆包含專利分類號。在實際實施上,專利資料庫110可以是各國家/地區的專利專責機構所設置的專利資料庫,也可以是民間單位、組織或個人所自行建立的專利資料庫,假設是自行建立的專利資料庫,其中儲存的專利文件可直接向各國家/地區的專利專責機構定期購買及更新。特別要說明的是,每一篇專利文件所標註的專利分類號,係專利專責機關的專業審查委員在核准專利申請之前,所界定的完整技術方案的總結技術元素,也就是說,是透過審查委員的專業總結出所涉及的不同技術組合。因此,只要得知專利分類號,無須瀏覽整篇專利便可快速且精確地得知專利所屬的技術領域。以美國專利公告號US 9,038,127為例,其所屬技術為資訊安全,特別是針對政策及防止未經授權使用數據(包含防止盜版、侵犯隱私或未經授權的數據修改),因此,審查委員將其專利分類號標註為「726/1」及「726/26」,用以與其所屬技術相互呼應,故每一個專利分類號皆可視為一個單獨的技術元素,當專利文件同時具有多個專利分類號時,可視為該專利由多個技術元素組合而成。The following describes the recommendation system and analysis method of the patent predictable and invalid comparison of the present invention with reference to the drawings. Please refer to "Picture 1" first. "Picture 1" is the system block of the patent predictable recommendation system of the present invention. In the figure, this system includes: a
檢索模組120用以提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫110以進行專利檢索,以便查詢出符合檢索條件的專利文件。在實際實施上,使用者鍵入的檢索條件可包含關鍵字(如:單字、專利分類號、公告號等等)、邏輯運算子(如:「AND」、「OR」、「NOT」等等)及指定檢索欄位(如:「@TI」、「/TTL」等等)。舉例來說,所述檢索條件可為:「物聯網 AND A63F 13/32」、「(網路)@TI」、「TTL/network」等等。特別要說明的是,不同的專利資料庫110可能使用不同的方式來指定檢索欄位,例如:以「@」或「/」來指定檢索欄位,其中,以中文專利資料庫為例,假設檢索條件為「(網路)@TI」,其代表將關鍵字「網路」的指定檢索欄位設為標題;以英文專利資料庫為例,假設檢索條件為「TTL/network」,其代表將關鍵字「network」的指定檢索欄位設為標題。另外,所述專利分類號可包含美國專利分類號(U.S. Patent Classification, UPC)、國際專利分類號(International Patent Classification, IPC)、合作專利分類號(Cooperative Patent Classification, CPC)及日本FI-F-Term等等,並且可具有大類(Class)及小類(Subclass)等層級。The
分析模組130用以載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度。在實際實施上,所述關聯規則演算法可為應用在資料探勘的Apriori演算法,並且同時搭配多維分析或時序分析以對載入的專利文件的專利分類號進行分析。具體而言,Apriori演算法是挖掘高頻項目集之布林值關聯規則中最具代表性的演算法,隨後發展的不同關聯規則演算法大多以Apriori演算法為基礎。其主要概念是在大量的資料集(如:專利文件)中,利用項目集(如:專利分類號)來建立關聯規則,並計算毎一個候選項目出現的數目,依據所設定的最小支持度為門檻,來衡量候選項目的關聯規則是否顯著。舉例來說,假設有4筆專利文件,每一筆專利文件包含的專利分類號以字母示意如下:The
專利文件一,其包含專利分類號A、C、D。Patent document one, which contains patent classification numbers A, C, D.
專利文件二,其包含專利分類號B、C、E。Patent document two, which contains patent classification numbers B, C, E.
專利文件三,其包含專利分類號A、B、C、E。Patent document three, which contains patent classification numbers A, B, C, E.
專利文件四,其包含專利分類號B、E。Patent document four, which contains patent classification numbers B and E.
在使用Apriori演算法建立關聯規則時,會進行高頻項目集之集合的搜尋與刪除,其步驟如下:When the Apriori algorithm is used to establish association rules, the search and deletion of the collection of high-frequency item sets is performed. The steps are as follows:
(1)將資料轉換為代碼或布林值表示的離散資料,在以累進搜尋的方式,從基層的單項專利分類號組合開始建立1-項目集之集合,經第一次掃描後可得C1並計算出各項目集所對應之支持度(以此例而言,1-項目集:{A}至{E},其對應之支持度依序為:0.5、0.75、0.75、0.25、0.75)。接下來比較所得之支持度與所定之支持度門檻S來決定高頻項目集,假設支持度門檻S為0.5,那麼項目集{D}將因為其支持度僅0.25而被排除,故得到高頻1-項目集有{A}、{B}、{C}及{E},將其記為L1。(1) Convert the data into discrete data expressed by code or Bollinger value, in a progressive search method, create a 1-item set from the combination of single patent classification numbers at the grassroots level, and C1 can be obtained after the first scan And calculate the corresponding support for each item set (in this case, 1-item set: {A} to {E}, the corresponding support degrees are in order: 0.5, 0.75, 0.75, 0.25, 0.75) . Next, compare the obtained support with the specified support threshold S to determine the high-frequency project set. Assuming the support threshold S is 0.5, then the project set {D} will be excluded because its support is only 0.25, so the high frequency is obtained. 1- The item set has {A}, {B}, {C}, and {E}, which is denoted as L1.
(2)將高頻1-項目集組合成6個2-項目集並記為C2;接著,同樣計算其支持度(以此例而言,2-項目集:{A, B}、{A, C}、{A, E}、{B, C}、{B, E}、{C,E},其對應之支持度依序為:0.25、0.5、0.25、0.5、0.75、0.5)。接著,同樣根據支持度門檻S決定高頻項目集,排除支持度為0.25的項目集{A, B}及{A, E},得到高頻2-項目集有{A, C}、{B, C}、{B, E}、{C,E},將其記為L2。(2) Combine the high-frequency 1-item sets into 6 2-item sets and record them as C2; then, also calculate their support (in this case, 2-item sets: {A, B}, {A , C}, {A, E}, {B, C}, {B, E}, {C,E}, the corresponding support levels are: 0.25, 0.5, 0.25, 0.5, 0.75, 0.5). Then, also determine the high-frequency item set according to the support threshold S, excluding the item sets {A, B} and {A, E} with a support degree of 0.25, and get the high-frequency 2-item set with {A, C}, {B , C}, {B, E}, {C,E}, and record it as L2.
(3)繼續累進搜尋,確認包含三個項目之項目集是否亦符合高頻項目集之特性,由於L2中各項目集在累進搜尋後,僅能找到一個3-項目集,即{B, C, E},故將其記為C3。此處,因為項目集{A, C, E}中的子項目集{A, E}並非高頻項目集,所以不須將項目集{A, C, E}列於C3中,而項目集{B, C, E}之子項目集{B, C}、{B, E}、{C,E}皆為高頻項目集,所以項目集{B, C, E}亦有機會成為高頻項目集。接著,計算出其支持度為0.5後,由於未低於支持度門檻S,故得到高頻3-項目集為{B, C, E},並記為L3。(3) Continue the progressive search to confirm whether the item set containing three items also meets the characteristics of the high-frequency item set. Since each item set in L2 can only find one 3-item set after the progressive search, that is, {B, C , E}, so record it as C3. Here, because the sub-project set {A, E} in the project set {A, C, E} is not a high-frequency project set, it is not necessary to list the project set {A, C, E} in C3, and the project set The sub-project sets {B, C}, {B, E}, {C, E} of {B, C, E} are all high-frequency project sets, so the project set {B, C, E} also has the opportunity to become a high-frequency Itemsets. Next, after calculating the support degree of 0.5, because it is not lower than the support threshold S, the high-frequency 3-item set is {B, C, E}, and is recorded as L3.
(4)接著,利用找到的高頻3-項目集{B, C, E}來建立關聯規則,在此例中可建立12種可能的關聯規則,並依序計算這些規則所對應之支持度與提昇度,如下表所示:
其中,支持度(Support)代表前項(X)和後項(Y)同時出現的概率,其數學式表示為:,表示所有資料集;提昇度(Lift)是置信度(Confidence)與後項支持度的比,大於1則意味著X的出現對Y的出現有促進作用,其數學式表示為:。Among them, Support (Support) represents the probability that the preceding term (X) and the following term (Y) appear at the same time, and its mathematical formula is expressed as: , Represents all data sets; Lift is the ratio of Confidence to the support of the latter term, greater than 1 means that the appearance of X promotes the appearance of Y, and its mathematical formula is expressed as: .
接下來,可根據支持度及提升度至少其中之一,從中找出顯著的關聯規則(如:支持度大於0.5或提昇度大於1),並且將顯著的關聯規則之關聯規則強度均設為強(或稱之為強連結),以及將非顯著的關聯規則之關聯規則強度均設為弱(或稱之為弱連結)。換句話說,所述關聯規則強度可依據查詢出的專利文件的數量、同時存在相應關聯規則所包含的專利分類號之專利文件數量等等來產生相應的強度,假設專利文件的數量為1024筆,關聯規則包含的專利分類號為「705」及「2」,則此關聯規則的關聯規則強度可計算在這1024筆專利文件中,每一筆專利文件同時存在專利分類號「705」及「2」的數量有多少,數量越多代表關聯規則強度越強,反之數量越少代表關聯規則強度越弱,也就是說,同一關聯規則所包含的專利分類號的組合,其同時出現在專利文件中的筆數與關聯規則強度成正相關。Next, according to at least one of the support degree and the promotion degree, a significant association rule can be found from it (eg, support degree is greater than 0.5 or promotion degree is greater than 1), and the strength of the association rule of the significant association rule is set to strong (Or called strong links), and set the strength of association rules of non-significant association rules to be weak (or called weak links). In other words, the strength of the association rule can be generated according to the number of searched patent documents, the number of patent documents with the patent classification number included in the corresponding association rule, and so on, assuming that the number of patent documents is 1024 , The patent classification numbers included in the association rules are "705" and "2", then the strength of the association rule of this association rule can be calculated in these 1024 patent documents, and each patent document has the patent classification numbers "705" and "2" "How many are there? The greater the number, the stronger the association rule strength. Conversely, the smaller the number, the weaker the association rule strength. That is, the combination of patent classification numbers included in the same association rule appears in the patent document at the same time. The number of strokes is positively correlated with the strength of association rules.
要補充說明的是,在實現Apriori演算法時,技術元素的關聯沒有一般商場購物籃分析的前項與後項之分,其關聯的項目都是實現方案的手段,沒有先後之分,除非是明確設定研發人員熟悉的技術元素為前項,來窺探要關聯哪一種技術元素為後項的推論(提昇度越高的關聯規則越好,因為其意味著前項的出現對後項的出現有促進作用),因此,以上例而言,可將「若B則C」及「若C則B」視為同一條關聯規則;將「若B則E」及「若E則B」視為同一條關聯規則;以及將「若C則E」及「若E則C」視為同一條關聯規則,總共得到9種可能的關聯規則。另外,若要針對某一專利做無效推論的證據查找的話,則要選擇關聯規則強度為強的關聯規則。反之,若要針對某一技術的創新元素做蒐集的話,則離群值的關聯規則(或稱之為分群的關聯規則)的可視化就變得極有意義,因為在龐大專利數據無法人工審閱的情況下,可以直觀的探索可組合的異業元素,這是在以往的商場購物籃分析不被採納的分析方法,因為在傳統的關聯規則分析中,這些關聯規則被視為雜訊(Noise)而排除。It should be added that when implementing the Apriori algorithm, the correlation of technical elements does not have the distinction between the antecedent and the antecedent of the general shopping basket analysis. The associated items are all means of implementing the plan, and there is no order, unless it is clear Set the technical elements familiar to the R&D personnel as the antecedent, to spy on which technical element to associate with the inferior (the higher the promotion, the better the association rules, because it means that the appearance of the antecedent promotes the appearance of the antecedent) Therefore, in the above example, "if B then C" and "if C then B" can be regarded as the same association rule; "if B then E" and "if E then B" can be regarded as the same association rule ; And "If C then E" and "If E then C" are regarded as the same association rule, a total of 9 possible association rules are obtained. In addition, if you want to search for evidence of invalidity inference for a patent, you need to choose an association rule with a strong association rule strength. On the contrary, if you want to collect the innovative elements of a certain technology, the visualization of the outlier association rules (or grouping association rules) becomes extremely meaningful, because in the case of huge patent data that cannot be manually reviewed Next, you can intuitively explore composable elements of different industries. This is an analysis method that has not been adopted in the previous shopping basket analysis, because in the traditional association rule analysis, these association rules are regarded as noise. exclude.
處理模組140用以將關聯規則強度為弱的關聯規則中的專利分類號進行組合以輸出為衍生專利建議。舉例來說,假設關聯規則強度為弱的關聯規則中,其包含的專利分類號為「E03D」及「H05K」,那麼,可將這二個專利分類號的組合作為衍生專利建議,換句話說,衍生專利建議中可以建議研發者在專利分類號「E03D」及「H05K」所各自代表的技術之組合基礎上,思考相關的技術或進一步改良的技術手段,此方式容易引導研發者,發想出具有可專利性的技術手段,因為關聯規則強度為弱,代表結合這二個技術的專利文件較少,所以在此基礎上進行技術發想比較不會與先前技術重複。另一方面,專利審查委員在進行專利審查時,也不容易找到可以用來核駁申請的對比前案,所以能夠有效提升專利獲准的機率。在實際實施上,所述衍生專利建議可嵌入與組合後的專利分類號相符的專利文件,例如:複製專利文件並合併至衍生專利建議,或以超連結方式將專利文件的號碼、名稱及儲存路徑嵌入衍生專利建議。The
另外,在實際實施上,本發明的系統更可包含建立模組,用以將每一關聯規則的專利分類號作為檢索條件,以便自專利資料庫110下載相符的專利文件,並且根據不同專利分類號對這些專利文件進行分類及儲存以形成技術元素庫。換句話說,每一個專利分類號所對應的技術皆可視為技術元素,所述技術元素庫中包含多個技術元素,每一個技術元素又具有對應的專利文件。在實際實施上,所述技術元素庫會將每一種技術所屬的前案專利文件分別收納在固定的資料夾(Folder)之中,例如:以專利分類號作為資料夾名稱。如此一來,在爾後需要參考類似的技術元素的前提下,可以直接在不同定義的資料夾中搜尋所有應用元素的不同技術手段與不同的應用場景,而不需要再從專利資料庫110反覆檢索或浪費其它的調研工作。In addition, in actual implementation, the system of the present invention may further include a module for using the patent classification number of each association rule as a search condition, so as to download matching patent files from the
接著,請參閱「第2圖」,「第2圖」為本發明可專利預測之分析方法的方法流程圖,其步驟包括:在專利資料庫110中儲存專利文件,每一專利文件皆包含專利分類號(步驟210);提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫110以進行專利檢索,查詢出符合檢索條件的專利文件(步驟220);載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度(步驟230);以及將關聯規則強度為弱的關聯規則中的專利分類號進行組合以輸出為衍生專利建議(步驟240)。透過上述步驟,即可透過載入與檢索條件相符的專利文件,並且以關聯規則演算法對載入的專利文件進行分析,用以建立包含專利分類號及關聯規則強度的關聯規則,以及從關聯規則強度為弱/強的關聯規則中,將其包含的專利分類號進行組合以輸出成為衍生專利建議/專利無效推論建議。Next, please refer to "Figure 2". "Figure 2" is a flow chart of the method of the patent predictive analysis method of the present invention. The steps include: storing patent documents in the
另外,在步驟240之後,還可將每一關聯規則的專利分類號作為檢索條件,用以自專利資料庫110下載相符的專利文件,並且根據不同專利分類號進行分類及儲存以形成技術元素庫(步驟250)。In addition, after
接著,請參閱「第3圖」,「第3圖」為本發明專利無效比對之推薦系統的系統方塊圖,此系統包含:專利資料庫310、檢索模組320、分析模組330及處理模組340。其中,專利資料庫310、檢索模組320及分析模組330分別與前述「第1圖」中的專利資料庫110、檢索模組120及分析模組130相同,故在此不再多做贅述,至於處理模組340與「第1圖」所示意的處理模組140之差異,兩者僅在於前者是從關聯規則強度為強的關聯規則中,將其包含的專利分類號進行組合以輸出為專利無效推論建議,而後者則是從關聯規則強度為弱的關聯規則中,將其包含的專利分類號進行組合以輸出為衍生專利建議。Next, please refer to "Figure 3", which is a system block diagram of a recommendation system for invalidation comparison of invention patents. This system includes:
接下來,請參閱「第4圖」,「第4圖」為本發明專利無效比對之分析方法的方法流程圖,其步驟包括:在專利資料庫310中儲存專利文件,每一專利文件皆包含專利分類號(步驟410);提供鍵入檢索條件,並且將鍵入的檢索條件傳送至專利資料庫310以進行專利檢索,查詢出符合檢索條件的專利文件(步驟420);載入查詢出的專利文件,並且以關聯規則演算法分析載入的專利文件的專利分類號,以及根據分析結果建立關聯規則,每一關聯規則包含至少二個專利分類號及一個關聯規則強度(步驟430);以及將關聯規則強度為強的關聯規則中的專利分類號進行組合以輸出為專利無效推論建議(步驟440)。在「第4圖」所示意的步驟440與前述「第2圖」所示意的步驟240之差異在於步驟 440是從關聯規則強度為強的關聯規則中,將其包含的專利分類號進行組合以輸出為專利無效推論建議,而後者則是從關聯規則強度為弱的關聯規則中,將其包含的專利分類號進行組合以輸出為衍生專利建議。至於步驟440之後,同樣可將每一關聯規則的專利分類號作為檢索條件,用以自專利資料庫310下載相符的專利文件,並且根據不同專利分類號進行分類及儲存以形成技術元素庫(步驟450)。Next, please refer to "Figure 4". "Figure 4" is a flowchart of a method for analyzing the invalidation comparison of invention patents. The steps include: storing patent documents in the
以下配合「第5圖」及「第6圖」以實施例的方式進行如下說明,請先參閱「第5圖」,「第5圖」為應用本發明產生衍生專利建議之示意圖。假設研發者為虛擬實境(Virtual Reality, VR)或擴增實境(Augmented Reality, AR)的技術背景,並且欲在此技術的基礎上進行創新發想。研發者可在輸入區塊511中鍵入檢索條件,如:「ACLM/"Virtual Reality"」或「ACLM/" Augmented Reality "」。此時,檢索模組120會將研發者鍵入的檢索條件傳送至專利資料庫110進行專利檢索,並查詢出符合的專利文件。接著,分析模組130從專利資料庫110載入這些被查詢出的專利文件,並且使用關聯規則演算法,如:Apriori演算法,針對這些專利文件的專利分類號進行關聯分析。在實際實施上,由於專利分類號為多階層資料,所以可以僅針對單一階層,如:大類,或是同時針對多階層,如:大類及小類(或稱之為次類),用以分別進行巨觀或微觀的關聯分析。以針對大類為例,使用關聯規則演算法分析後,可產生相應的關聯規則,所述關聯規則可以圖形化方式呈現在第一顯示區塊521,其中,線條兩端是關聯規則中相關聯的專利分類號(在此例中為大類),而線條的粗細則代表關聯規則強度,舉例來說,粗線條代表高度關聯,也就是說粗線條兩端的大類,其對應的關聯規則強度為強,同時亦代表這兩個大類是常被運用的技術元素組合。另外,除了上述單獨分析大類之外,在實際實施上,亦可使用相同方式同時分析大類及小類,用以產生相應的關聯規則,並且同樣以圖形化的方式呈現在第二顯示區塊522。值得一提的是,在第二顯示區塊522中出現許多分群的關聯規則,例如:「709/227、709/217」、「705/26.1、705/27.1、705/2」、「703/2、703/1」等等,這些分群的關聯規則可以視為「分群創新元素關聯規則」,也就是說,這些關聯規則中的專利分類號,其代表的技術是非常適合作為被結合的技術元素(例如:適合異業結合的技術元素)。最後,處理模組140會將關聯規則強度為弱的關聯規則中的專利分類號進行組合以輸出為衍生專利建議,其輸出方式可為建立檔案或直接顯示在建議區塊530。此時,研發者即可瀏覽建議區塊530中顯示的衍生專利建議,例如,從中發想如何結合虛擬實境、電腦與數位處理系統的多電腦傳輸,特別是遠端資料存取(美國專利分類號:709/217)及電腦至電腦的會話/連接建立(美國專利分類號:709/227)等技術,以便產出具有可專利性的技術。在研發者發想的過程中,研發者還可同時在建議區塊530中點選顯示的專利公告號,以便開啟對應的專利文件進行瀏覽。要補充說明的是,當涉及的技術元素(專利分類號;或稱之為項目)數量過多時,還可嘗試依專利文件公告的時間先後(技術發展進程),分別列為不同區段,由最近公告(即:第一區段)拆分至最早公告(第n區段)來加以分析且使用圖形化方式呈現,例如,第一區(1~100筆)、第二區(101~200筆)、第三區(201~300筆)、並以此類推至第n區。如此一來,即可窺探技術元素在不同時間區間(如:發展期、成熟期及衰退期)的發展及其運用情況。The following description will be made in conjunction with "figure 5" and "figure 6" by way of example. Please refer to "figure 5" first. "figure 5" is a schematic diagram of applying the present invention to generate derivative patent proposals. Assume that the developer is a virtual reality (Virtual Reality, VR) or augmented reality (Augmented Reality, AR) technical background, and want to innovate on the basis of this technology. The developer can enter the search criteria in the
如「第6圖」所示意,「第6圖」為應用本發明產生專利無效推論建議之示意圖。假設研發者遭遇到專利侵權訴訟或警告,可直接將系爭專利的專利分類號作為檢索條件鍵入輸入區塊611,此時,檢索模組320會將研發者鍵入的檢索條件傳送至專利資料庫310進行專利檢索,並查詢出符合的專利文件。接著,分析模組330從專利資料庫310載入這些被查詢出的專利文件,並且使用關聯規則演算法,如:Apriori演算法,針對這些專利文件的專利分類號進行關聯分析,並且根據分析結果產生關聯規則。所述關聯規則以圖形化方式呈現在顯示區塊620,其中,線條兩端是關聯規則中相關聯的專利分類號,而線條的粗細則代表關聯規則強度,舉例來說,粗線條代表高度關聯(即:關聯規則強度為強),反之則代表低度關聯(即:關聯規則強度為弱)。至此,上述流程與「第5圖」的流程大同小異,差別僅在於是否同時分析大類及小類,在「第6圖」中為了簡化說明僅分析大類。然而,接下來處理模組340會從關聯規則強度為強的關聯規則中,將其包含的專利分類號進行組合以輸出為專利無效推論建議,其輸出方式可為建立檔案或直接顯示在建議區塊630。此時,研發者即可瀏覽建議區塊630中顯示的專利無效推論建議,從而得知更多與系爭專利直接相關的技術元素組合及其對應的專利文件,以此例而言,假設系爭專利的專利分類號為345/619,代表系爭專利所屬技術領域為計算機圖形處理和選擇性視覺顯示系統中的圖形操作,而從專利無效推論建議中可得知其與影像分析技術(大類為382)結合的數量最多,因此可推論與系爭專利直接相關的技術元素組合為影像分析技術,所以在查找對比前案時,可將影像分析技術作為限縮檢索範圍的依據,能夠精準找到具有高度關聯性的前案專利文件,用以作為舉發系爭專利時的證據與論述上的辯證支持。換句話說,有別於「第5圖」的處理模組140針對關聯規則強度為弱的關聯規則,「第6圖」的處理模組340是針對關聯規則強度為強的關聯規則,因為關聯規則強度為強的關聯規則,代表存在關聯規則所包含的專利分類號的專利文件之數量也越多,故容易從中找到對比的前案專利文件,有利於後續作為舉發系爭專利的證據與論述上的辯證支持,進而提高撤銷系爭專利之專利權的機率。As shown in "Figure 6", "Figure 6" is a schematic diagram of the application of the present invention to generate a suggestion for invalidation of patents. Assuming that the developer encounters a patent infringement lawsuit or warning, he can directly enter the patent classification number of the disputed patent as the search condition into the
綜上所述,可知本發明與先前技術之間的差異在於透過載入與檢索條件相符的專利文件,並且以關聯規則演算法對載入的專利文件進行分析,用以建立包含專利分類號及關聯規則強度的關聯規則,以及從關聯規則強度為弱/強的關聯規則中,將其包含的專利分類號進行組合以輸出成為衍生專利建議/專利無效推論建議,藉由此一技術手段可以解決先前技術所存在的問題,進而達成提高可專利預測及無效比對的便利性之技術功效。In summary, it can be seen that the difference between the present invention and the prior art lies in that by loading patent documents that match the search conditions, and analyzing the loaded patent documents with an association rule algorithm, it is used to establish a patent classification number and Association rules with strong association rules, and from association rules with weak or strong association rules, combine the patent classification numbers they contain to output as derivative patent suggestions/invalidation inference suggestions, which can be solved by a technical means The problems of the prior art, and then achieve the technical efficiency of improving the convenience of patentable prediction and invalidation comparison.
雖然本發明以前述之實施例揭露如上,然其並非用以限定本發明,任何熟習相像技藝者,在不脫離本發明之精神和範圍內,當可作些許之更動與潤飾,因此本發明之專利保護範圍須視本說明書所附之申請專利範圍所界定者為準。Although the present invention is disclosed as the foregoing embodiments, it is not intended to limit the present invention. Any person who is familiar with similar arts can make some changes and modifications without departing from the spirit and scope of the present invention. The scope of patent protection shall be subject to the definition of the scope of patent application attached to this specification.
110、310‧‧‧專利資料庫 120、320‧‧‧檢索模組 130、330‧‧‧分析模組 140、340‧‧‧處理模組 150、350‧‧‧建立模組 511、611‧‧‧輸入區塊 521‧‧‧第一顯示區塊 522‧‧‧第二顯示區塊 530、630‧‧‧建議區塊 620‧‧‧顯示區塊 步驟210‧‧‧在至少一專利資料庫中儲存多個專利文件,每一專利文件皆包含至少一專利分類號 步驟220‧‧‧提供鍵入一檢索條件,並且將鍵入的該檢索條件傳送至所述專利資料庫以進行專利檢索,查詢出符合該檢索條件的所述專利文件 步驟230‧‧‧載入查詢出的所述專利文件,並且以一關聯規則演算法分析載入的所述專利文件的所述專利分類號,以及根據分析結果建立多個關聯規則,每一關聯規則包含至少二個所述專利分類號及一關聯規則強度 步驟240‧‧‧將該關聯規則強度為弱的所述關聯規則中的所述專利分類號進行組合以輸出為一衍生專利建議 步驟250‧‧‧將每一關聯規則的所述專利分類號作為該檢索條件,用以自所述專利資料庫下載相符的所述專利文件,並且根據不同所述專利分類號進行分類及儲存以形成一技術元素庫 步驟410‧‧‧在至少一專利資料庫中儲存多個專利文件,每一專利文件皆包含至少一專利分類號 步驟420‧‧‧提供鍵入一檢索條件,並且將鍵入的該檢索條件傳送至所述專利資料庫以進行專利檢索,查詢出符合該檢索條件的所述專利文件 步驟430‧‧‧載入查詢出的所述專利文件,並且以一關聯規則演算法分析載入的所述專利文件的所述專利分類號,以及根據分析結果建立多個關聯規則,每一關聯規則包含至少二個所述專利分類號及一關聯規則強度 步驟440‧‧‧將該關聯規則強度為強的所述關聯規則中的所述專利分類號進行組合以輸出為一專利無效推論建議 步驟450‧‧‧將每一關聯規則的所述專利分類號作為該檢索條件,用以自所述專利資料庫下載相符的所述專利文件,並且根據不同所述專利分類號進行分類及儲存以形成一技術元素庫 110, 310‧‧‧ Patent database 120, 320‧‧‧ search module 130、330‧‧‧Analysis module 140, 340‧‧‧ processing module 150、350‧‧‧Build module 511, 611‧‧‧ input block 521‧‧‧The first display block 522‧‧‧Second display block 530, 630‧‧‧ suggested block 620‧‧‧Display block Step 210‧‧‧ Store multiple patent documents in at least one patent database, each patent document contains at least one patent classification number Step 220 ‧‧ provide to enter a search condition, and transmit the entered search condition to the patent database for patent search, and search out the patent file that meets the search condition Step 230‧‧‧ Load the searched out patent document, and analyze the patent classification number of the loaded patent document with an association rule algorithm, and establish multiple association rules according to the analysis result, each association The rules include at least two of the patent classification numbers and an association rule strength Step 240 ‧‧‧ The patent classification number in the association rule with weak association rule strength is combined to output as a derivative patent proposal Step 250: Use the patent classification number of each association rule as the search condition to download the matching patent file from the patent database, and classify and store it according to different patent classification numbers Form a library of technical elements Step 410‧‧‧ Store multiple patent documents in at least one patent database, each patent document contains at least one patent classification number Step 420‧‧‧ Provide to enter a search condition, and transmit the entered search condition to the patent database for patent search, and search out the patent file that meets the search condition Step 430‧‧‧ Load the searched out patent document, and analyze the patent classification number of the loaded patent document with an association rule algorithm, and establish a plurality of association rules according to the analysis result, each association The rules include at least two of the patent classification numbers and an association rule strength Step 440‧‧‧ The patent classification number in the association rule with strong association rule strength is combined to output as a patent invalidation inference suggestion Step 450: Use the patent classification number of each association rule as the search condition to download the matching patent file from the patent database, and classify and store according to different patent classification numbers Form a library of technical elements
第1圖為本發明可專利預測之推薦系統的系統方塊圖。 第2圖為本發明可專利預測之分析方法的方法流程圖。 第3圖為本發明專利無效比對之推薦系統的系統方塊圖。 第4圖為本發明專利無效比對之分析方法的方法流程圖。 第5圖為應用本發明產生衍生專利建議之示意圖。 第6圖為應用本發明產生專利無效推論建議之示意圖。Figure 1 is a system block diagram of the patent predictable recommendation system of the present invention. Figure 2 is a flow chart of a method for analyzing the patentable prediction of the present invention. Fig. 3 is a system block diagram of a recommendation system for invalidation comparison of invention patents. Figure 4 is a flow chart of the method for analyzing the invalidity comparison of invention patents. Figure 5 is a schematic diagram of applying the present invention to generate derivative patent proposals. Fig. 6 is a schematic diagram of applying the present invention to produce suggestions for invalidation of patents.
110‧‧‧專利資料庫 110‧‧‧ Patent database
120‧‧‧檢索模組 120‧‧‧Search module
130‧‧‧分析模組 130‧‧‧Analysis module
140‧‧‧處理模組 140‧‧‧ processing module
150‧‧‧建立模組 150‧‧‧Create module
Claims (16)
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW107121551A TWI696923B (en) | 2018-06-22 | 2018-06-22 | Recommendation system for patentable prediction and invalid comparison and analysis method thereof |
CN201910522353.0A CN110633264B (en) | 2018-06-22 | 2019-06-17 | Research and development auxiliary system and method using patent database |
US16/446,637 US20190391976A1 (en) | 2018-06-22 | 2019-06-20 | Research and development auxiliary system using patent database and method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW107121551A TWI696923B (en) | 2018-06-22 | 2018-06-22 | Recommendation system for patentable prediction and invalid comparison and analysis method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
TW202001602A TW202001602A (en) | 2020-01-01 |
TWI696923B true TWI696923B (en) | 2020-06-21 |
Family
ID=69942282
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW107121551A TWI696923B (en) | 2018-06-22 | 2018-06-22 | Recommendation system for patentable prediction and invalid comparison and analysis method thereof |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI696923B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI789681B (en) * | 2021-01-14 | 2023-01-11 | 廖光陽 | Patent matrix system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101334788A (en) * | 2008-07-18 | 2008-12-31 | 袁祖六 | Patent classification method |
CN103824161A (en) * | 2014-02-28 | 2014-05-28 | 惠州学院 | Cloud computing patentability technology navigation system and method |
TWM529875U (en) * | 2016-03-04 | 2016-10-01 | Univ Dayeh | Patent classification searching system by disassembling data field |
-
2018
- 2018-06-22 TW TW107121551A patent/TWI696923B/en active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101334788A (en) * | 2008-07-18 | 2008-12-31 | 袁祖六 | Patent classification method |
CN103824161A (en) * | 2014-02-28 | 2014-05-28 | 惠州学院 | Cloud computing patentability technology navigation system and method |
TWM529875U (en) * | 2016-03-04 | 2016-10-01 | Univ Dayeh | Patent classification searching system by disassembling data field |
Non-Patent Citations (3)
Title |
---|
劉可瑜,專利文件分類檢索之改良,東吳大學資訊管理學系碩士論文,上架日:2012年2月15日 * |
徐明立,專利趨勢變化探勘之研究,國立交通大學資訊管理研究所碩士論文,公開日期:2004年 * |
徐明立,專利趨勢變化探勘之研究,國立交通大學資訊管理研究所碩士論文,公開日期:2004年。 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI789681B (en) * | 2021-01-14 | 2023-01-11 | 廖光陽 | Patent matrix system |
Also Published As
Publication number | Publication date |
---|---|
TW202001602A (en) | 2020-01-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | Consisrec: Enhancing gnn for social recommendation via consistent neighbor aggregation | |
Yan et al. | Data mining in the construction industry: Present status, opportunities, and future trends | |
Carrera-Rivera et al. | How-to conduct a systematic literature review: A quick guide for computer science research | |
Seeliger et al. | Semantic web technologies for explainable machine learning models: A literature review. | |
Wan et al. | Aminer: Search and mining of academic social networks | |
JP7503000B2 (en) | System and method for investigating relationships between entities - Patents.com | |
Soibelman et al. | Management and analysis of unstructured construction data types | |
Hao et al. | Knowledge map-based method for domain knowledge browsing | |
US11074275B2 (en) | Automatically propagating tagging of content items in a content management system environment | |
US20210019665A1 (en) | Machine Learning Model Repository Management and Search Engine | |
CN110633264B (en) | Research and development auxiliary system and method using patent database | |
Im et al. | Linked tag: image annotation using semantic relationships between image tags | |
JP2012528406A (en) | Merging search results | |
US20150324350A1 (en) | Identifying Content Relationship for Content Copied by a Content Identification Mechanism | |
WO2014100459A2 (en) | Systems and methods for using non-textual information in analyzing patent matters | |
Jaillant et al. | Unlocking digital archives: cross-disciplinary perspectives on AI and born-digital data | |
CA2786355C (en) | Associative memory technology for analysis of requests for proposal | |
Cao | [Retracted] Mode Optimization and Rule Management of Intellectual Property Rights Protection of Educational Resource Data Based on Machine Learning Algorithm | |
Weikum et al. | Knowledge harvesting: achievements and challenges | |
TWI696923B (en) | Recommendation system for patentable prediction and invalid comparison and analysis method thereof | |
Anand et al. | Explainable information retrieval | |
Garba et al. | Federated search techniques: an overview of the trends and state of the art | |
US20080301082A1 (en) | Knowledge base comprising executable stories | |
US11283839B2 (en) | Enforcement knowledge graph-based data security rule change analysis | |
US10754969B2 (en) | Method to allow for question and answer system to dynamically return different responses based on roles |