UpKG: A Framework to Insert New Domains in Knowledge Graphs

Abstract

In recent years, the creation of Knowledge Graphs (KGs) has advanced significantly. They have become essential in several domains, such as e-commerce. E-commerce applications apply them in the search and recommendation of products, question-answering systems, and virtual chatbots, among other tasks. However, e-commerce must constantly cover new domains/categories to respond to new user needs. This process requires rigorous analysis to cover new domains, adding novel knowledge not stored in the KG under use. This MS.c Dissertation proposes, builds, and evaluates a framework we named UpKG to insert new domains in existing KGs within an e-commerce context. Our approach relies on questions and answers collected. We applied our study in the real-world context of the GoBots, company specialized in e-commerce solutions in Latin America, to facilitate the new domain insertion process in a KG. The conducted case study was on an existing KG, whose triples were dominated by the Automobile domain. Our investigation by applying the UpKG framework covered and inserted new triplets related to the Appliances domain, obtaining a KG that supports the Automobiles and Appliances domain. The final KG was populated by 3382 new instances, extracted from 1338 real-world questions and answers from the GoBots database. In addition, we carried out two types of evaluations: the first was focused on evaluating the ontology, and the second on evaluating the KG. To evaluate the ontology, we explored three tools: Reasoner Pellet, OOPS!, and OntoDebug, which allows us to ensure the consistency and coherence of the modeled ontology. To evaluate the KG, 30 Competency Questions were used (15 from the Automobiles domain and 15 from the Appliances domain), which allowed us to evaluate whether the instances align with the application, which is to answer compatibility questions. The evaluation showed the viability and applicability of the UpKG framework in different domains within e-commerce. Our contribution allows the expansion of the KG to new domains.

Esta investigación propone, construye y evalúa un framework que llamamos UpKG para insertar nuevos dominios en KG existentes dentro de un contexto de e-commerce. Aplicamos nuestro framework en el contexto del mundo real de GoBots, empresa especializada en soluciones e-commerce. El caso de estudio realizado se centró en un KG existente, cuyas triplas estaban dominados por el sector del automóvil. Nuestra investigación, aplicando el framework UpKG, cubrió e insertó nuevas triplas relacionados con el dominio Electrodomésticos, obteniendo un KG que admite el dominio Automóviles y Electrodomésticos. El KG final estuvo poblado por 3382 nuevas instancias, extraídas de 1338 preguntas y respuestas del mundo real de la base de datos GoBots. Además, realizamos dos tipos de evaluaciones: la primera estuvo enfocada a evaluar la ontología y la segunda a evaluar el KG. Para evaluar la ontología, exploramos tres herramientas: Reasoner Pellet, OOPS! y OntoDebug, que nos permite garantizar la consistencia y coherencia de la ontología modelada. Para evaluar el KG se utilizaron 30 Preguntas de Competencia (15 del dominio Automóviles y 15 del dominio Electrodomésticos), lo que permitió evaluar si las instancias responden preguntas de compatibilidad. La evaluación mostró la viabilidad y aplicabilidad del framework UpKG en diferentes ámbitos del e-commerce.