Evaluating BERT-based Scientific Relation Classifiers for Scholarly Knowledge Graph Construction on Digital Library Collections

The rapid growth of research publications has placed great demands on digital libraries (DL) for advanced information management technologies. To cater to these demands, techniques relying on knowledge-graph structures are being advocated. In such graph-based pipelines, inferring semantic relations between related scientific concepts is a crucial step. Recently, BERT-based pre-trained models have been popularly explored for automatic relation classification. Despite significant progress, most of them were evaluated in different scenarios, which limits their comparability. Furthermore, existing methods are primarily evaluated on clean texts, which ignores the digitization context of early scholarly publications in terms of machine scanning and optical character recognition (OCR). In such cases, the texts may contain OCR noise, in turn creating uncertainty about existing classifiers' performances. To address these limitations, we started by creating OCR-noisy texts based on three clean corpora. Given these parallel corpora, we conducted a thorough empirical evaluation of eight Bert-based classification models by focusing on three factors: (1) Bert variants; (2) classification strategies; and, (3) OCR noise impacts. Experiments on clean data show that the domain-specific pre-trained Bert is the best variant to identify scientific relations. The strategy of predicting a single relation each time outperforms the one simultaneously identifying multiple relations in general. The optimal classifier's performance can decline by around 10% to 20% in F-score on the noisy corpora. Insights discussed in this study can help DL stakeholders select techniques for building optimal knowledge-graph-based systems.

Improving Scholarly Knowledge Representation: Evaluating BERT-based Models for Scientific Relation Classification

With the rapidly growing number of research publications, there is a vast amount of scholarly information that needs to be organized in digital libraries. To deal with this challenge, digital libraries use semantic techniques to build knowledge-base structures for organizing scientific information. Identifying relations between scientific terms can help with the construction of a representative knowledge-based structure. While advanced automated techniques have been developed for relation extraction, many of these techniques were evaluated under different scenarios, which limits their comparability. To this end, this study presents a thorough empirical evaluation of eight Bert-based classification models by exploring two factors: 1) Bert model variants, and 2) classification strategies. To simulate real-world settings, we conduct our sentence-level assessment using the abstracts of scholarly publications in three corpora, two of which are distinct corpora and the third of which is the union of the first two. Our findings show that SciBert models perform better than Bert-BASE models. The strategy of classifying a single relation each time is preferred in the corpus consisting of abundant scientific relations, while the strategy of identifying multiple relations at one time is beneficial to the corpus with sparse relations. Our results offer recommendations to the stakeholders of digital libraries for selecting the appropriate technique to build a structured knowledge-based system for the ease of scholarly information organization.