The Annotated Corpus of Classical Tibetan (ACTib) - Version 2.0 (Segmented & POS-tagged)

This corpus consisting of >185 million tokens is a segmented and part-of-speech tagged version of

Wallman, Jeff, Rowinski, Zach, Ngawang Trinley, Tomlinson, Chris, & Keutzer, Kurt. (2017). Collection of Tibetan etexts compiled by the Buddhist Digital Resource Center [Data set]. Zenodo. http://doi.org/10.5281/zenodo.821218

using the training data of

Hill, Nathan W., & Garrett, Edward. (2017). A part-of-speech (POS) tagged corpus of Classical Tibetan [Data set]. Zenodo. http://doi.org/10.5281/zenodo.574878

The code for segmenting and POS tagging any Tibetan file can be found on GitHub.

This Version 2 of ACTib is based on the same XML files as ACTib Version 1 (http://doi.org/10.5281/zenodo.823707), but contains both segmented and POS-tagged files and is improved in a number of ways, although post-processing was still done automatically and no manual correction was involved. For details of this improved annotation method see:

Meelen, Marieke, Roux, Élie & Hill, Nathan (forthcoming). 'Optimisation of the largest annotated Tibetan corpus combining rule-based, memory-based & deep-learning methods' in TALLIP.