Dual Input Stream Transformer for Vertical Drift Correction in Eye-Tracking Reading Data
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE Transactions on Pattern Analysis and Machine Intelligence
Volume: 46
Issue: 12
Pages: 8715-8726
eISSN: 1939-3539
ISSN: 0162-8828
DOI: 10.1109/TPAMI.2024.3411938
Abstract:We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17%. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field.
https://eprints.bournemouth.ac.uk/40059/
Source: Scopus
Dual Input Stream Transformer for Vertical Drift Correction in Eye-Tracking Reading Data.
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE Trans Pattern Anal Mach Intell
Volume: 46
Issue: 12
Pages: 8715-8726
eISSN: 1939-3539
DOI: 10.1109/TPAMI.2024.3411938
Abstract:We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17%. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field.
https://eprints.bournemouth.ac.uk/40059/
Source: PubMed
Dual Input Stream Transformer for Vertical Drift Correction in Eye-Tracking Reading Data
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
Volume: 46
Issue: 12
Pages: 8715-8726
eISSN: 1939-3539
ISSN: 0162-8828
DOI: 10.1109/TPAMI.2024.3411938
https://eprints.bournemouth.ac.uk/40059/
Source: Web of Science (Lite)
Dual Input Stream Transformer for Vertical Drift Correction in Eye-Tracking Reading Data.
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE Trans. Pattern Anal. Mach. Intell.
Volume: 46
Pages: 8715-8726
https://eprints.bournemouth.ac.uk/40059/
Source: DBLP
Dual Input Stream Transformer for Vertical Drift Correction in Eye-Tracking Reading Data.
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE transactions on pattern analysis and machine intelligence
Volume: 46
Issue: 12
Pages: 8715-8726
eISSN: 1939-3539
ISSN: 0162-8828
DOI: 10.1109/tpami.2024.3411938
Abstract:We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17%. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field.
https://eprints.bournemouth.ac.uk/40059/
Source: Europe PubMed Central
Dual input stream transformer for vertical drift correction in eye-tracking reading data.
Authors: Mercier, T.M., Budka, M., Vasilev, M.R., Kirkby, J.A., Angele, B. and Slattery, T.J.
Journal: IEEE Transactions on Pattern Analysis and Machine Intelligence
Volume: 46
Issue: 12
Pages: 8715-8726
ISSN: 0162-8828
Abstract:We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17 %. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field.
https://eprints.bournemouth.ac.uk/40059/
Source: BURO EPrints