Unstructured to Visualized: Transforming Data into Knowledge Graphs via Processing Pipeline
Afifa Fatima, D. Naga Jyothi, Prerana Jalapally · 2024
As the amount of unstructured data increases at a significant rate, the task of extracting useful and valuable data becomes challenging. In this paper, two novel state-of-the-art pipelines for relation extraction from unstructured textual data are presented using ML and NLP approaches. Each pipeline has three modules including coreference resolution, named entity linking, and relationship extraction. This study is based on a Netflix dataset, the relations derived from this dataset are then used to build a recommendation system. The first pipeline employs rule-based RE, while the second implements a sophisticated REBEL model to extract relations. After evaluating their performance and efficacy, the best of the pipeline is used to create a movie-based recommendation system employing Neo4j to store relations and similarity scored from the description.