OryzaGP: rice gene and protein dataset for named-entity recognition

Area of Science:

• Plant Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Text mining is crucial for extracting biological entities like genes and proteins from scientific literature.
• Limited datasets and benchmarks exist for plant molecular biology, particularly for rice, hindering advanced analysis.
• Accurate named-entity recognition for rice is challenging due to the scarcity of specialized resources.

Purpose of the Study:

• To develop a new benchmark dataset for rice gene and protein name recognition.
• To facilitate the application of advanced machine learning methods for analyzing rice literature.
• To establish a shared task for evaluating text mining approaches in rice molecular biology.

Main Methods:

• Compiled a dataset of titles and abstracts from PubMed focusing on rice research.
• Utilized the 5th Biomedical Linked Annotation Hackathon for data sharing via PubAnnotation.
• Designed the dataset to serve as a benchmark for named-entity recognition tasks.

Main Results:

• Created a novel, publicly available dataset specifically for rice gene/protein entity extraction.
• The dataset addresses the lack of benchmarks for rice literature analysis.
• Facilitates comparative evaluation of different text mining and machine learning approaches.

Conclusions:

• The new rice dataset is essential for advancing text mining in plant molecular biology.
• It enables the development and evaluation of improved named-entity recognition systems for rice.
• Promotes open comparison and progress in analyzing rice scientific literature through shared tasks.