Screening early stage ideas in technology development processes: a text mining and k-nearest neighbours approach using patent information

Abstract

Applying previous idea screening approaches to large amounts of early stage ideas is recognised as challenging since they rely heavily on manual tasks and human judgments. Considering that technological factors are more important than others in early phases of technology development processes, we propose a machine learning approach to screening ideas by linking the contents of ideas implied in patented inventions and the technological value of the ideas. At the heart of the proposed approach are the text mining technique, to construct keyword vectors from patents, and the k-nearest neighbours algorithm, to capture the relationships between the keyword vectors and the numbers of forward citations of the patents. Integration of these methods makes it possible to assess large amounts of early stage ideas in terms of their potential technological value. A case study of pharmaceutical technology shows that our approach is useful for filtering out ideas of little technological value.