TY - GEN
T1 - On Applicability of Neural Language Models for Readability Assessment in Filipino
AU - Ibañez, Michael
AU - Reyes, Lloyd Lois Antonie
AU - Sapinit, Ranz
AU - Hussien, Mohammed Ahmed
AU - Imperial, Joseph Marvin
PY - 2022/7/27
Y1 - 2022/7/27
N2 - In the field of automatic readability assessment (ARA), the current trend in the research community focuses on the use of large neural language models such as BERT as evidenced from its high performance in other downstream NLP tasks. In this study, we dissect the BERT model and applied it to readability assessment in a low-resource setting using a dataset in the Filipino language. Results show that extracting embeddings separately from various layers of BERT obtain relatively similar performance with models trained using a diverse set of handcrafted features and substantially better than using conventional transfer learning approach.
AB - In the field of automatic readability assessment (ARA), the current trend in the research community focuses on the use of large neural language models such as BERT as evidenced from its high performance in other downstream NLP tasks. In this study, we dissect the BERT model and applied it to readability assessment in a low-resource setting using a dataset in the Filipino language. Results show that extracting embeddings separately from various layers of BERT obtain relatively similar performance with models trained using a diverse set of handcrafted features and substantially better than using conventional transfer learning approach.
KW - BERT
KW - Neural language models
KW - Readability assessment
UR - http://www.scopus.com/inward/record.url?scp=85135895276&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-11647-6_118
DO - 10.1007/978-3-031-11647-6_118
M3 - Chapter in a published conference proceeding
AN - SCOPUS:85135895276
SN - 9783031116469
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 573
EP - 576
BT - Artificial Intelligence in Education. Posters and Late Breaking Results, Workshops and Tutorials, Industry and Innovation Tracks, Practitioners’ and Doctoral Consortium - 23rd International Conference, AIED 2022, Proceedings
A2 - Rodrigo, Maria Mercedes
A2 - Matsuda, Noburu
A2 - Cristea, Alexandra I.
A2 - Dimitrova, Vania
PB - Springer Science and Business Media Deutschland GmbH
T2 - 23rd International Conference on Artificial Intelligence in Education, AIED 2022
Y2 - 27 July 2022 through 31 July 2022
ER -