On the Importance of Signer Overlap for Sign Language Detection

Sign language detection, identifying if someone is signing or not, is becoming crucially important for its applications in remote conferencing software and for selecting useful sign data for training sign language recognition or translation tasks. We argue that the current benchmark data sets for sign language detection estimate overly positive results that do not generalize well due to signer overlap between train and test partitions. We quantify this with a detailed analysis of the effect of signer overlap on current sign detection benchmark data sets. Comparing accuracy with and without overlap on the DGS corpus and Signing in the Wild, we observed a relative decrease in accuracy of 4.17% and 6.27%, respectively. Furthermore, we propose new data set partitions that are free of overlap and allow for more realistic performance assessment. We hope this work will contribute to improving the accuracy and generalization of sign language detection systems.

Clean Text and Full-Body Transformer: Microsoft's Submission to the WMT22 Shared Task on Sign Language Translation

This paper describes Microsoft's submission to the first shared task on sign language translation at WMT 2022, a public competition tackling sign language to spoken language translation for Swiss German sign language. The task is very challenging due to data scarcity and an unprecedented vocabulary size of more than 20k words on the target side. Moreover, the data is taken from real broadcast news, includes native signing and covers scenarios of long videos. Motivated by recent advances in action recognition, we incorporate full body information by extracting features from a pre-trained I3D model and applying a standard transformer network. The accuracy of the system is further improved by applying careful data cleaning on the target text. We obtain BLEU scores of 0.6 and 0.78 on the test and dev set respectively, which is the best score among the participants of the shared task. Also in the human evaluation the submission reaches the first place. The BLEU score is further improved to 1.08 on the dev set by applying features extracted from a lip reading model.