Patent retrieval based on semantic search

Abstract

    Keyword-based search is a popular, classical, and limited search method. In contrast, semantic search have the ability to understand the specific topic or meaning of each paragraph and sentence. For the above reasons, a new search method was proposed to solve the problem of finding similar patents because it may prevent the registration of the same ideas and protect the ownership and commercial rights. Two main techniques in semantic search include: vector search and Natural Language Processing for English patent documents. The data source was taken from the USPTO website belonging to the USA government. The new outcome of this research is: to find documents that are close to the meaning of a given document more accurately and rapidly.

The initial results of the present study showed that 10 most similar patents were found in a database of 694 patents within only 0.55692 seconds. This result is much faster than other similar search methods.