Concept-based patent image retrieval

Abstract Recently, the intellectual property and information retrieval communities have shown increasing interest in patent image retrieval, which could further enhance the current practices of patent search. In this context, this article presents an approach for automatically extracting concept information describing the patent image content to support searchers during patent retrieval tasks. The proposed approach is based on a supervised machine learning framework, which relies upon image and text analysis techniques. Specifically, we extract textual and visual low-level features from patent images and train detectors, which are capable of identifying global concepts in patent figures. To evaluate this approach we have selected a dataset from the footwear domain and trained the concept detectors with different feature combinations. The results of the experiments show that the combination of textual and visual information of patent images demonstrates the best performance outperforming both single visual and textual features results. The outcome of this experiment provides a first evidence that concept detection can be applied in the domain of patent image retrieval and could be integrated in existing real world applications to support patent searching.