The WLASL (Li et al., 2020) is a large video dataset for Word-Level American Sign Language recognition, available for download. It brings together many different sets of ASL videos that had been shared publicly.

However, there is a serious lack of consistency in the gloss labeling associated with signs across the multiple datasets. See: http://www.bu.edu/asllrp/rpt21/asllrp21.pdf.

We provide gloss labels for a large subset of the WLASL data consistent with the glossing conventions for the ASLLRP Sign Bank, with the following aims in mind:

(1) These annotations can increase the value of the WLASL data for sign recognition research, by virtue of the consistency in the labels attached to the signs in these collections.

(2) This also makes it possible to put data from the WLASL and the ASLLRP Sign Bank together, to create an even larger and richer resource than either of these data sources on their own.

These annotations can be used with the data that can be downloaded from the WLASL site.