在專利申請的過程中,需要一個專利代理人來委託其專利申請的事宜,在智慧局所認可的專利代理人中要找出一個對於自己要申請的專利較適合的專利代理人是不容易的,智慧局並無一個介面可供申請專利者查詢代理人過去所代理的專利。而本研究希望能利用近三年的專利文件,找尋出專利代理人所代理過的專利,並從這些專利中依據國際分類號統計出各專利代理人的專利領域為何,再者利用專利文件中的內容、特性找尋出代表此專利文件的關鍵字後,將專利代理人的專利領域與資訊擷取中的四種空間向量模型以比對出原專利文件中的專利代理人為目的來找出哪一種空間向量的推薦效果較佳,並視為本推薦代理人機制的推薦人選之一。本研究發現在資訊擷取中餘弦係數的推薦成效較好,且在推薦原代理人排名前30的條件之下達到五成以上的正確率。
During the patent application process, most of the applicants or inventors will entrust a patent agent whom is authorized by TIPO to handle the application. And it is difficult to find out a most suitable attorney to coordinate with every patent applying case. TIPO do not provide an interface to search cases that patent agent had deal with in the past for patent applicants. In this study collect nearly three years’ patent document, classify each agent’s patent case, and then according to the International patent classifications to find out the patent agents' professional scopes. Also extract contains of the patent document mining the keywords. Take patent agents and patent information retrieval in the field of four vector space model to compare the original patent documents for the purpose of the patent agent to find out which one was better the recommendation vector space, and considered the recommended agent one mechanism of recommended candidates. The study found that in the information capture results in the cosine of the recommended good, and recommended the original agents in the top 30 under the conditions to achieve over half of the correct rate.