Patent Technical Function-effect Representation and Mining Method

Weidong Liu, Piying Zhang, Wenbo Qiao

Abstract: With increasing global competition of intellectual property, a large number of unstructured patent texts are generated for technology protection. The ocean of patent texts include many long sentences about technologies, technical functions, technical effects and complexity relations between them, which make it difficult to textual representation and mining. To solve the above issues, we represent a patent by its technical function-effects, which are mined from the patent. The model represents functions/effects by valence utility-technologies and represents text by association relations between functions and effects. We evaluate our model by comparing with the state-of-the-art models on the patent data set. The results show that our model outperforms other models in evaluation measurement. Such representation can be applied to patent information retrieval and patent text analysis.