TY - JOUR
T1 - Large Scale Subject Category Classification of Scholarly Papers With Deep Attentive Neural Networks
AU - Kandimalla, Bharath
AU - Rohatgi, Shaurya
AU - Wu, Jian
AU - Giles, C. Lee
N1 - Publisher Copyright:
Copyright © 2021 Kandimalla, Rohatgi, Wu and Giles.
PY - 2020
Y1 - 2020
N2 - Subject categories of scholarly papers generally refer to the knowledge domain(s) to which the papers belong, examples being computer science or physics. Subject category classification is a prerequisite for bibliometric studies, organizing scientific publications for domain knowledge extraction, and facilitating faceted searches for digital library search engines. Unfortunately, many academic papers do not have such information as part of their metadata. Most existing methods for solving this task focus on unsupervised learning that often relies on citation networks. However, a complete list of papers citing the current paper may not be readily available. In particular, new papers that have few or no citations cannot be classified using such methods. Here, we propose a deep attentive neural network (DANN) that classifies scholarly papers using only their abstracts. The network is trained using nine million abstracts from Web of Science (WoS). We also use the WoS schema that covers 104 subject categories. The proposed network consists of two bi-directional recurrent neural networks followed by an attention layer. We compare our model against baselines by varying the architecture and text representation. Our best model achieves micro- (Formula presented.) measure of 0.76 with (Formula presented.) of individual subject categories ranging from 0.50 to 0.95. The results showed the importance of retraining word embedding models to maximize the vocabulary overlap and the effectiveness of the attention mechanism. The combination of word vectors with TFIDF outperforms character and sentence level embedding models. We discuss imbalanced samples and overlapping categories and suggest possible strategies for mitigation. We also determine the subject category distribution in CiteSeerX by classifying a random sample of one million academic papers.
AB - Subject categories of scholarly papers generally refer to the knowledge domain(s) to which the papers belong, examples being computer science or physics. Subject category classification is a prerequisite for bibliometric studies, organizing scientific publications for domain knowledge extraction, and facilitating faceted searches for digital library search engines. Unfortunately, many academic papers do not have such information as part of their metadata. Most existing methods for solving this task focus on unsupervised learning that often relies on citation networks. However, a complete list of papers citing the current paper may not be readily available. In particular, new papers that have few or no citations cannot be classified using such methods. Here, we propose a deep attentive neural network (DANN) that classifies scholarly papers using only their abstracts. The network is trained using nine million abstracts from Web of Science (WoS). We also use the WoS schema that covers 104 subject categories. The proposed network consists of two bi-directional recurrent neural networks followed by an attention layer. We compare our model against baselines by varying the architecture and text representation. Our best model achieves micro- (Formula presented.) measure of 0.76 with (Formula presented.) of individual subject categories ranging from 0.50 to 0.95. The results showed the importance of retraining word embedding models to maximize the vocabulary overlap and the effectiveness of the attention mechanism. The combination of word vectors with TFIDF outperforms character and sentence level embedding models. We discuss imbalanced samples and overlapping categories and suggest possible strategies for mitigation. We also determine the subject category distribution in CiteSeerX by classifying a random sample of one million academic papers.
UR - http://www.scopus.com/inward/record.url?scp=85177851007&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85177851007&partnerID=8YFLogxK
U2 - 10.3389/frma.2020.600382
DO - 10.3389/frma.2020.600382
M3 - Article
AN - SCOPUS:85177851007
SN - 2504-0537
VL - 5
JO - Frontiers in Research Metrics and Analytics
JF - Frontiers in Research Metrics and Analytics
M1 - 600382
ER -