UWSpace is currently experiencing technical difficulties resulting from its recent migration to a new version of its software. These technical issues are not affecting the submission and browse features of the site. UWaterloo community members may continue submitting items to UWSpace. We apologize for the inconvenience, and are actively working to resolve these technical issues.
 

KG-Pipeline: An Automated Knowledge Graph Generation Framework

Thumbnail Image

Files

Vezvaei_Alireza.pdf (1.74 MB)

Date

2022-09-16

Authors

Vezvaei, Alireza

Journal Title

Journal ISSN

Volume Title

Publisher

University of Waterloo

Abstract

Knowledge Graphs (KGs) have many applications, specifically in information retrieval and question answering. Community projects are conducted for building large-scale KGs with crowdsourcing, but building KGs with this approach is costly and sometimes infeasible. Considering the rapidly growing amount of unstructured text on the Web, we highly need systems for automatic KG generation. We propose KG-Pipeline, a general-purpose end-to-end pipeline designed for automatically constructing KGs from unstructured text documents. We leverage state-of-the-art NLP models for implementing various components of the pipeline. We also utilize our generated KGs in Question Answering (QA) and evaluate the performance of our system on a QA benchmark, comparing it to previous work and an information retrieval baseline model.

Description

Keywords

knowledge graph, question answering, information extraction, NLP, natural language processing

LC Keywords

Citation

URI

http://hdl.handle.net/10012/18750

Collections

Theses
Computer Science