AraT5: Text-to-Text Transformers for Arabic Language Generation
This work addresses the need for effective text-to-text transformers in Arabic, a language with diverse dialects, by providing improved models and a benchmark, though it is incremental as it builds on existing T5 frameworks.
The authors tackled the problem of applying multilingual T5 models to Arabic language generation by introducing AraT5 models and a new benchmark (ARGEN), achieving significantly better performance than mT5 on 52 out of 59 test sets and setting new SOTAs on ARGEN and ARLUE benchmarks.
Transfer learning with a unified Transformer framework (T5) that converts all language problems into a text-to-text format was recently proposed as a simple and effective transfer learning approach. Although a multilingual version of the T5 model (mT5) was also introduced, it is not clear how well it can fare on non-English tasks involving diverse data. To investigate this question, we apply mT5 on a language with a wide variety of dialects--Arabic. For evaluation, we introduce a novel benchmark for ARabic language GENeration (ARGEN), covering seven important tasks. For model comparison, we pre-train three powerful Arabic T5-style models and evaluate them on ARGEN. Although pre-trained with ~49 less data, our new models perform significantly better than mT5 on all ARGEN tasks (in 52 out of 59 test sets) and set several new SOTAs. Our models also establish new SOTA on the recently-proposed, large Arabic language understanding evaluation benchmark ARLUE (Abdul-Mageed et al., 2021). Our new models are publicly available. We also link to ARGEN datasets through our repository: https://github.com/UBC-NLP/araT5.