Semantic Structuring of E-commerce Texts: The QART Framework

Abstract

The challenge of transforming natural language texts into structured knowledge representations is important to enhance data integration in e-commerce. We developed the QART framework to address this challenge of converting e-commerce questions and answers into RDF triples for integration into existing Knowledge Graphs (KGs). The QART framework consists of four main steps: field selection and pre-processing, text-to-text conversion, text triplifying, and RDF triple curation. These steps aim to manage the volume and complexity of e-commerce data while ensuring semantic correctness and consistency with predefined ontologies. Our evaluations demonstrated that intermediary steps, such as text summarization, produce competitive results and can improve the quality of the resulting triples.