Detection and splitting of constructs of SAPPhIRE model to support automatic structuring of analogies
Sonal Keshwani, Amaresh Chakrabarti · 2017
The objective of this work is to structure a natural language description of analogies into a common causal language – which is chosen here to be SAPPhIRE model of causality. The motivation is to create a database of analogies that is structured so as to support focused search for analogies across the database. This should provide the benefit of utilizing the enormous data available on the Internet, while also providing relevant analogies to the designers as search results. This objective is achieved by implementing the following three steps: Firstly, detection of SAPPhIRE constructs in a document, achieved with an F-Measure of 0.834 using a text-classification approach; secondly, splitting sentences containing multiple SAPPhIRE constructs, achieved with an accuracy of 76.5% using a rule based approach; Thirdly, prediction of SAPPhIRE constructs for each text-input, implemented using the method proposed in literature. With these three steps, the time required to structure analogies into a common causal language can be reduced, thereby supporting population of the database and hence enabling designers in retrieving relevant analogies for novel idea generation.