Problems and Approaches in NLP-based Patent Analysis

Abstract

Recently technologies in Natural Language Processing (NLP) become more popular with their increasing performances in various fields. We discuss some problems and possible approaches of applying those technologies to patent analysis. We show that patent-specific domain knowledge is important to analyze patents. Several approaches are built on the domain knowledge and show high performance. Meanwhile, toward some general problems like lexical gap or low frequency words, interactive analysis with an analyzing tool in a cognitive way is more appropriate.