Summary of Is Child-directed Speech Effective Training Data For Language Models?, by Steven Y. Feng et al.
Is Child-Directed Speech Effective Training Data for Language Models?
by Steven Y. Feng, Noah D. Goodman, Michael C. Frank
First submitted to arxiv on: 7 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper investigates the features of child-directed speech that support language modeling objectives and whether these features are unique in supporting high performance. The authors train GPT-2 and RoBERTa models on 29M words of English child-directed speech, a synthetic dataset (TinyDialogues), and compare to OpenSubtitles, Wikipedia, and other datasets. They evaluate the syntactic and semantic knowledge of these models using developmentally-inspired evaluations. Pretraining experiments reveal that local discourse ordering affects model results, but global properties do not. The findings support the hypothesis that child language input is not uniquely valuable for training language models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at how children learn language so quickly, even with much less data than language models are trained on. It compares different types of data to see which one works best for training language models. The authors use two popular models (GPT-2 and RoBERTa) and test them on a special dataset made just for this study. They also compare these results to other datasets like movie subtitles, Wikipedia, and more. The surprising result is that the way the data is ordered, rather than the amount of data itself, affects how well the models perform. |
Keywords
» Artificial intelligence » Discourse » Gpt » Pretraining
