Eye-tracking-while-reading: A living survey of datasets with open library support

Area of Science:

• Cognitive Science
• Computational Linguistics
• Data Science

Background:

• Eye-tracking-while-reading corpora are crucial for understanding reading cognition and developing AI applications.
• Existing datasets are fragmented across disciplines, hindering interoperability and reuse due to a lack of data sharing standards.
• The increasing size and diversity of these datasets highlight the need for better organization and accessibility.

Purpose of the Study:

• To enhance transparency and clarity of existing eye-tracking-while-reading datasets across disciplines.
• To simplify the sharing and discovery of newly created datasets through a living online overview.
• To promote FAIR data principles and good scientific practices in eye-tracking research.

Main Methods:

• Compiling an extensive overview of over 55 features for existing eye-tracking-while-reading datasets.
• Publishing a dynamic online resource to facilitate the sharing of new datasets.
• Integrating publicly available datasets into the `pymovements` Python package for unified access.

Main Results:

• A comprehensive overview of eye-tracking-while-reading datasets with over 55 features is now available.
• A living online overview simplifies the submission and discovery of new datasets.
• The `pymovements` Python package provides integrated access to publicly available eye-tracking datasets.

Conclusions:

• This work strengthens the FAIR principles in eye-tracking-while-reading research.
• Improved data accessibility and interoperability will foster reproducibility and replication of studies.
• The initiative promotes standardized data sharing practices within the research community.