This repository comprises the collected dataset of handwritten transcriptions of the ASAP SAS.
This database may be used for non-commercial research purpose only. If you publish material based on this database - please refer to the following paper:
Christian Gold and Torsten Zesch. 2020. Exploring the Impact of Handwriting Recognition on the Automated Scoring of Handwritten Student Answers. International Conference on Frontiers in Handwriting Recognition (ICFHR).
The latest version of the dataset can be downloaded here. It is based on the Short Answer Scoring (SAS) dataset of the Automated Student Assessment Prize (ASAP). We used texts from the test set for our transcription, as we used the train set for Automatic Scorring.
It is planned to increase the dataset further by asking students participating a lecture to copy one sheet per lecture. We will update the dataset resulting in a new version, when new transcriptions are available. As from 2 prompts more than 150 transcriptions each could already be gathered, we continue to collect transcriptions from other prompts first before filling up the previous prompts. The text images presented are cut out of the alligned scans and have the standard size of 1960 × 1575 px. An information file per prompt has been added, containing all filenames, ASAP IDs, original texts, pen color, and the status.
With version 1.0 from 1st July 2020 a total of 350 transcriptions out of two prompts are collected. Prompt 3 - 185 texts Prompt 4 - 165 texts (full description file is missing)