利用學習排序法進行專利無效搜尋

Abstract

專利在今日扮演著重要的資源，擁有專利意味著擁有著相對應的排他權，可以避免被他人所抄襲、避免被侵犯的損失。而專利持有人也可以透過專利進行轉讓授權，進而獲取高額的授權金，或者是使用專利去進行侵權的告訴來賺取賠償金。而隨著專利的數量逐日上升，企業們越來越有可能會面臨到與專利相關的訴訟，因此提出專利無效訴訟來保護自身成為了很重要的事情。然而傳統的專利搜尋方式十分費工費時，需要大量的人工去做搜尋以及判讀的工作。本研究使用學習排序法來進行專利無效搜尋，將與目標專利相關性做排序，讓潛在的侵權專利排名上升，提供專利判讀人員更佳的專利排序組合，讓判讀人員可以省下更多的時間。

Patents are important resources nowadays. The inventor(s) or the assignee(s) who owns a patent has the exclusive right to prevent or exclude others from making, using, selling, offering to sell or importing the invention. Besides, firms can also use their patents to prevent their products from being infringed, license out their patents to gain royalty fees, or attack other firms via patent infringement lawsuits. As the number of patents granted increases annually, it is more likely that a firm will face patent infringement lawsuits. When a company is accused of patent infringement, to protect itself from this lawsuit, it is a common practice that the defendant files a patent invalidity lawsuit against the plaintiff. To achieve this goal, the defendant should conduct prior art search on the focal patent to find prior arts whose existence can invalid the focal patent. Thus, this study takes the learning to rank approach to design our patent invalidity search technique. Our empirical evaluation result shows that the effectiveness of our proposed technique and its enhanced version is satisfactory in supporting patent invalidity search.