Table 4:
Performance comparison of models using full texts and our implemented models using paper abstracts (and authors and research areas as categories for basis-customized models) as inputs in the AAPR data set.
ModelsAccuracy
using full text (Yang et al., 2018)
LSTM 60.5
MHCNN 67.7
using abstract and categories (our setting)
LSTM 60.6
MHCNN 63.7
BiLSTM 61.7
BiLSTM +word-basis-cust 65.8
BiLSTM +attention-basis-cust 65.9
BiLSTM +encoder-basis-cust 66.2
ModelsAccuracy
using full text (Yang et al., 2018)
LSTM 60.5
MHCNN 67.7
using abstract and categories (our setting)
LSTM 60.6
MHCNN 63.7
BiLSTM 61.7
BiLSTM +word-basis-cust 65.8
BiLSTM +attention-basis-cust 65.9
BiLSTM +encoder-basis-cust 66.2
Close Modal