From Babble to Words: Pre-Training Language Models on Continuous Streams of Phonemes
This work addresses a niche problem for researchers in computational linguistics and language acquisition, offering incremental insights into phoneme-based training methods.
The paper tackled the problem of evaluating language models trained on phoneme streams instead of text by developing a pipeline to convert text datasets to phonemes and applying it to pre-training and benchmarks. The result showed that phoneme-based training slightly reduces performance on traditional tasks but offers analytical and practical benefits, with specific numbers not provided.
Language models are typically trained on large corpora of text in their default orthographic form. However, this is not the only option; representing data as streams of phonemes can offer unique advantages, from deeper insights into phonological language acquisition to improved performance on sound-based tasks. The challenge lies in evaluating the impact of phoneme-based training, as most benchmarks are also orthographic. To address this, we develop a pipeline to convert text datasets into a continuous stream of phonemes. We apply this pipeline to the 100-million-word pre-training dataset from the BabyLM challenge, as well as to standard language and grammatical benchmarks, enabling us to pre-train and evaluate a model using phonemic input representations. Our results show that while phoneme-based training slightly reduces performance on traditional language understanding tasks, it offers valuable analytical and practical benefits.