專利知識本體之建立與推薦系統

隨著網路的普及化及各國對專利權的重視，許多發明家及企業都會想為自己研發的產品申請專利，避免被他人抄襲及生產商品等，來保護自己的所研發的商品或研究的結果。然而近年來環保意識抬頭，各國也逐漸將紙本專利文件轉為數位化管理專利文件，目前專利文件領域中尚未建立一個較為完整的知識本體（Ontology），過去有關專利文件檢索之研究鮮少採用基於文件分析之本體設計方法，整合應用於專利文件的知識領域上，本研究以語意網作為基礎再結合推薦系統，對於致力專利文件創作的人而言，提供了可以快速檢索未來要撰寫的專利之相關文件以及從檢索結果，獲得更多的相關推薦專利文件，讓專利創作者可以從所獲取的資訊，了解未來專利文件創作的方向及避免專利侵權的事情發生。本研究利用語意網作為基礎，將專利檢索和推薦系統做結合，讓使用者可以在搜索專利的同時不再是只有一項專利文件可以參考，而是會有許多相關的專利文件可以參考撰寫，推薦系統的協同過濾會根據專利文件創作者的偏好和歷史的查詢紀錄來給予推薦，以利專利文件創作者獲取更多相關的專利文件資訊，進而創作更好的專利文件。

關鍵字

專利檢索；推薦系統；語意網；智慧檢索

並列摘要

With the popularization of the network and the importance of patent rights in various countries, many inventors and enterprises will want to patent their own products, to avoid being copied by others and the production of goods, etc., to protect their research and development of goods or research results. However, in recent years, environmental consciousness has increased, and countries have gradually turned paper-based patent documents into digital management patent documents. The field of patent documents has not yet established a more complete patent ontology for retrieval. In the past, on the search of patent documents rarely use the ontology based on document analysis design methods to integrate the knowledge areas applied to patent documents. In this study, we combine the recommendation system with scheme of the semantic web. The person who is committed to the creation of patent documents can quickly retrieve the relevant documents of the patent to be written in the future and obtain more relevant recommendation patent documents from the search results, so that the patent can understand the direction of future patent document creation and avoid patent infringement. This study uses Semantic Web as a basis to combine patent search and recommendation systems so that users can search for patents while no longer having only one patent file to refer to, but there will be many relevant patent documents to be recommend. The collaborative filtering of the recommendation system will be recommended according to the patent document author's preferences and historical query records, so that the patent document creator can obtain more relevant patent document information, and thus create a better patent document.