Automatic prior art searching

In the intellectual property field two task s are of high relevance: prior art search ing and patent classification. Prior art search is fundamental for many strategic issues such as patent granting, freedom to operate and opposition . Accurate classification of patent documents according to the IPC code system is vital for the interoperability between different patent offices and for the prior art search task involved in a patent application procedure. In this paper, we report our experiments with prior art search ing and patent classification in the context of CLEF - IP ’10 evaluation track. In the Prior Art Candidates search task, we strongly improved our last year’s model based on our experiments on training data (MAP 0.22) , but official results , alas, w ere far from the expected one s (MAP 0.14) .

According to EPO, i t is estimated that 80% of the knowledge is found in patent documents. Due to its importance as source of knowledge and to the delay in patent analysis caused the growth of applications, new areas of knowledge and size of patent databases, new tools to automate patent search ing and classification process es have become a hot topic in the last decades . As example, we can cite the challenges CLEF 2009, TREC - CHEM 2009 - 2010 and the workshops SIGIR 20 00, ACL 2003 and NTCIR 3 - 8 which all have tasks dedicated to patent retrieval. In that context, the CLEF - IP 2010 evaluation track proposes two tasks for automation of prior art search ing and of patent classification.