English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We present MatSci-NLP, a natural language benchmark for evaluating theperformance of natural language processing (NLP) models on materials sciencetext. We construct the benchmark from publicly available materials science textdata to encompass seven different NLP tasks, including conventional NLP taskslike named entity recognition and relation classification, as well as NLP tasksspecific to materials science, such as synthesis action retrieval which relatesto creating synthesis procedures for materials. We study various BERT-basedmodels pretrained on different scientific text corpora on MatSci-NLP tounderstand the impact of pretraining strategies on understanding materialsscience text. Given the scarcity of high-quality annotated data in thematerials science domain, we perform our fine-tuning experiments with limitedtraining data to encourage the generalize across MatSci-NLP tasks. Ourexperiments in this low-resource training setting show that language modelspretrained on scientific text outperform BERT trained on general text. MatBERT,a model pretrained specifically on materials science journals, generallyperforms best for most tasks. Moreover, we propose a unified text-to-schema formultitask learning on \benchmark and compare its performance with traditionalfine-tuning methods. In our analysis of different training methods, we findthat our proposed text-to-schema methods inspired by question-answeringconsistently outperform single and multitask NLP fine-tuning methods. The codeand datasets are publicly available at\url{https://github.com/BangLab-UdeM-Mila/NLP4MatSci-ACL23}.

MatSci-NLP: Evaluating Scientific Language Models on Materials Science Language Tasks Using Text-to-Schema Modeling