Details
| Original language | English |
|---|---|
| Title of host publication | Lernen, Wissen, Daten, Analysen 2018 |
| Subtitle of host publication | Proceedings of the Conference "Lernen, Wissen, Daten, Analysen" |
| Pages | 239-250 |
| Number of pages | 12 |
| Publication status | Published - 2018 |
| Event | 2018 Conference "Learning, Knowledge, Data, Analytics", LWDA 2018 - Mannheim, Germany Duration: 22 Aug 2018 → 24 Aug 2018 |
Publication series
| Name | CEUR Workshop Proceedings |
|---|---|
| Publisher | CEUR Workshop Proceedings |
| Volume | 2191 |
| ISSN (Print) | 1613-0073 |
Abstract
Handwriting recognition approaches usually use the semantic context of the individual letters. This helps to achieve highly accurate classification results. As it is less likely to detect semantically invalid letters, these approaches imply an error correction. In most cases, this is a beneficial side-effect but it makes the approaches not usable if these semantic invalid letters should be detected, e.g., in the case of educational software where spelling mistakes should be recognized. For this purpose, we developed an advanced context-free handwriting recognition which is based on multiple techniques from the field of deep learning. We motivate the individual components of our approach and show how the components can benefit from each other. In the experimental section, the behavior of our new approach is analyzed in detail and the classification accuracy is compared between the different approaches.
Keywords
- Deep learning, Educational software, Handwriting recognition, Human-machine interface, Learning support
ASJC Scopus subject areas
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Lernen, Wissen, Daten, Analysen 2018: Proceedings of the Conference "Lernen, Wissen, Daten, Analysen". 2018. p. 239-250 (CEUR Workshop Proceedings; Vol. 2191).
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Multi-Stage Deep Learning for Context-Free Handwriting Recognition
AU - Lückehe, Daniel
AU - Mühlpforte, Nicole
AU - Effenberg, Alfred O.
AU - Von Voigt, Gabriele
PY - 2018
Y1 - 2018
N2 - Handwriting recognition approaches usually use the semantic context of the individual letters. This helps to achieve highly accurate classification results. As it is less likely to detect semantically invalid letters, these approaches imply an error correction. In most cases, this is a beneficial side-effect but it makes the approaches not usable if these semantic invalid letters should be detected, e.g., in the case of educational software where spelling mistakes should be recognized. For this purpose, we developed an advanced context-free handwriting recognition which is based on multiple techniques from the field of deep learning. We motivate the individual components of our approach and show how the components can benefit from each other. In the experimental section, the behavior of our new approach is analyzed in detail and the classification accuracy is compared between the different approaches.
AB - Handwriting recognition approaches usually use the semantic context of the individual letters. This helps to achieve highly accurate classification results. As it is less likely to detect semantically invalid letters, these approaches imply an error correction. In most cases, this is a beneficial side-effect but it makes the approaches not usable if these semantic invalid letters should be detected, e.g., in the case of educational software where spelling mistakes should be recognized. For this purpose, we developed an advanced context-free handwriting recognition which is based on multiple techniques from the field of deep learning. We motivate the individual components of our approach and show how the components can benefit from each other. In the experimental section, the behavior of our new approach is analyzed in detail and the classification accuracy is compared between the different approaches.
KW - Deep learning
KW - Educational software
KW - Handwriting recognition
KW - Human-machine interface
KW - Learning support
UR - http://www.scopus.com/inward/record.url?scp=85053614097&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85053614097
T3 - CEUR Workshop Proceedings
SP - 239
EP - 250
BT - Lernen, Wissen, Daten, Analysen 2018
T2 - 2018 Conference "Learning, Knowledge, Data, Analytics", LWDA 2018
Y2 - 22 August 2018 through 24 August 2018
ER -