학술논문
MasakhaPOS: Part-of-Speech Tagging for Typologically Diverse African Languages
Document Type
Working Paper
Author
Dione, Cheikh M. Bamba; Adelani, David; Nabende, Peter; Alabi, Jesujoba; Sindane, Thapelo; Buzaaba, Happy; Muhammad, Shamsuddeen Hassan; Emezue, Chris Chinenye; Ogayo, Perez; Aremu, Anuoluwapo; Gitau, Catherine; Mbaye, Derguene; Mukiibi, Jonathan; Sibanda, Blessing; Dossou, Bonaventure F. P.; Bukula, Andiswa; Mabuya, Rooweither; Tapo, Allahsera Auguste; Munkoh-Buabeng, Edwin; Koagne, victoire Memdjokam; Kabore, Fatoumata Ouoba; Taylor, Amelia; Kalipe, Godson; Macucwa, Tebogo; Marivate, Vukosi; Gwadabe, Tajuddeen; Elvis, Mboning Tchiaze; Onyenwe, Ikechukwu; Atindogbe, Gratien; Adelani, Tolulope; Akinade, Idris; Samuel, Olanrewaju; Nahimana, Marien; Musabeyezu, Théogène; Niyomutabazi, Emile; Chimhenga, Ester; Gotosa, Kudzai; Mizha, Patrick; Agbolo, Apelete; Traore, Seydou; Uchechukwu, Chinedu; Yusuf, Aliyu; Abdullahi, Muhammad; Klakow, Dietrich
Source
Subject
Language
Abstract
In this paper, we present MasakhaPOS, the largest part-of-speech (POS) dataset for 20 typologically diverse African languages. We discuss the challenges in annotating POS for these languages using the UD (universal dependencies) guidelines. We conducted extensive POS baseline experiments using conditional random field and several multilingual pre-trained language models. We applied various cross-lingual transfer models trained with data available in UD. Evaluating on the MasakhaPOS dataset, we show that choosing the best transfer language(s) in both single-source and multi-source setups greatly improves the POS tagging performance of the target languages, in particular when combined with cross-lingual parameter-efficient fine-tuning methods. Crucially, transferring knowledge from a language that matches the language family and morphosyntactic properties seems more effective for POS tagging in unseen languages.
Comment: Accepted to ACL 2023 (Main conference)
Comment: Accepted to ACL 2023 (Main conference)