概念式中文專利搜尋系統之研究

Abstract

本論文目標在於實作出一個非傳統搜尋方法之專利搜尋系統：藉由概念式的中文描述語句，進行輸入查詢，以文意評分，提供使用者判斷及利用，整理混沌的專利資訊。對專利資料及查詢子句預先處理，並利用潛在語意分析(LSA, Latent Semantic Analysis)演算法進行專利文意相關度排序，提供使用者一個便利的檢索環境，並且讓檢索出來的資訊能夠最符合使用者的真切需求。同時利用LSA演算法各項參數調整，針對專利文獻做最佳化，增加結果的準確度與合理性。將專利資料庫發揮出最大的功效，進而幫助智慧財產權以及專利商品化的推廣與實施。 本論文附錄亦針對至目前為止，在專利相關領域上的研究方向及機構，作簡單的分析介紹。提供欲對智慧財產權作進一步了解者，可以更簡易、輕鬆地找到相關研究資源，提升我國人在智慧財產權上面的國際競爭力。

The main goal of this thesis is to implement a patent retrieval system with unconventional searching method. With the conceptual query sentence described in Chinese, it can score the results by similarity of their meanings and use the results to support the users. This system can provide the users a simple and convenient query environment, and let the information retrieved fit what users really want in order to marshalling the chaos patent information. By processing the query sentences and the patent information in advance, we use Latent Semantic Analysis algorithm to classify the information by their real meanings and show their relative degree to the query sentences. By means of adjusting parameters in LSA algorithm, we can optimize the system with patent documents and increase the accuracy and reasonability of results at the same time. This will enable the patent database to bring out most benefit and to help the intellectual property rights and patent commercialization spread and carry out eventually. This thesis also has a brief introduction on the patent related research areas and organizations at present day in the appendix. It can help people who want to get further understanding about intellectual property rights to find related research resources easily, which can also improve the international competitive ability on intellectual property rights for Taiwan researchers.