Ilmu Komputer & AI editorial
Open AccessOA2022
SCICERO: A Deep Learning and NLP Approach for Generating Scientific Knowledge Graphs in the Computer Science Domain
SCICERO is a novel approach that leverages NLP and deep learning Transformer models to automatically generate knowledge graphs from scientific papers, addressing the bottleneck of manual annotation in scholarly communication.
D. Dessรญ; Francesco Osborne; D. Recupero; D. Buscaldi; E. Mottaยท Knowledge-Based Systemsยท 2022ยท DOI 10.1016/j.knosys.2022.109945
The core problem
Science communication faces significant bottlenecks due to the increasing volume of published research papers and their non-machine-accessible, document-based format, which hinders efficient exploration, reading, and reuse of research outcomes. Knowledge Graphs (KGs) have emerged as a core technology to semantically interlink entities and curate scholarly information, making it machine-readable and understandable. However, the manual annotation required from researchers to populate these KGs is a major drawback. To overcome this, the authors propose SCICERO, a knowledge graph generation approach that automatically processes research article text to generate a KG of research entities.
Innovation
SCICERO employs Natural Language Processing (NLP) techniques to parse the content of scientific papers, discovering entities and relationships. It then utilizes state-of-the-art Deep Learning Transformer models to make sense of and validate the extracted information. Finally, it applies Semantic Web best practices to construct the knowledge graph. The approach integrates these components into a pipeline that takes raw text from research articles as input and outputs a structured KG.
Introduction
Science communication faces significant bottlenecks due to the increasing volume of published research papers and their non-machine-accessible, document-based format, which hinders efficient exploration, reading, and reuse of research outcomes. Knowledge Graphs (KGs) have emerged as a core technology to semantically interlink entities and curate scholarly information, making it machine-readable and understandable. However, the manual annotation required from researchers to populate these KGs is a major drawback. To overcome this, the authors propose SCICERO, a knowledge graph generation approach that automatically processes research article text to generate a KG of research entities.
Why it matters
The authors discuss the implications of SCICERO for scholarly communication, highlighting its potential to automate the curation of research outcomes and enable more efficient exploration and reuse of scientific knowledge. They also address limitations and future work, such as extending the approach to other domains and improving the robustness of the extraction and validation steps. The discussion emphasizes the importance of machine-readable scholarly information and the role of KGs in advancing science.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member contentโฆ