System for Identification of References
Vaishak Sajeev, Peeta Basa Pati · 2022 2nd Asian Conference on Innovation in Technology (ASIANCON) · 2022
Citations and references are made in almost every kind of document. Most documents tend to make direct or indirect reference to past events or articles. Such references help us obtain holistic picture of current events by picking up information from earlier works. Identifying these references improves the reading experience & provides a clear overall picture. Manual identification of such references and linkage creation for associated historical events is not only tedious but also a gross misutilization of human resources.This work reports a system that can identify & locate textual statements containing reference entities in an article. BERT is utilized to encode the sentence tokens which subsequently are classified with various standard and ensemble algorithms. Random Forest, an ensemble algorithm, at 92% accuracy & F1 score, is observed to be the most efficient in identifying the sentences with reference entities.